first commit

.gitignore

@@ -1,2 +1,13 @@
-.idea
-.idea/**
+DUMMY1
+DUMMY2
+DUMMY3
+logs
+__pycache__
+.ipynb_checkpoints
+.*.swp
+
+build
+*.c
+monotonic_align/monotonic_align
+/.vs/vits/FileContentIndex
+output.pth

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2021 Jaehyeon Kim
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,12 +1,58 @@
----
-title: Vits Nyaru
-emoji: 🌖
-colorFrom: indigo
-colorTo: blue
-sdk: gradio
-sdk_version: 3.1.4
-app_file: app.py
-pinned: false
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# VITS: Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech
+
+### Jaehyeon Kim, Jungil Kong, and Juhee Son
+
+In our recent [paper](https://arxiv.org/abs/2106.06103), we propose VITS: Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech.
+
+Several recent end-to-end text-to-speech (TTS) models enabling single-stage training and parallel sampling have been proposed, but their sample quality does not match that of two-stage TTS systems. In this work, we present a parallel end-to-end TTS method that generates more natural sounding audio than current two-stage models. Our method adopts variational inference augmented with normalizing flows and an adversarial training process, which improves the expressive power of generative modeling. We also propose a stochastic duration predictor to synthesize speech with diverse rhythms from input text. With the uncertainty modeling over latent variables and the stochastic duration predictor, our method expresses the natural one-to-many relationship in which a text input can be spoken in multiple ways with different pitches and rhythms. A subjective human evaluation (mean opinion score, or MOS) on the LJ Speech, a single speaker dataset, shows that our method outperforms the best publicly available TTS systems and achieves a MOS comparable to ground truth.
+
+Visit our [demo](https://jaywalnut310.github.io/vits-demo/index.html) for audio samples.
+
+We also provide the [pretrained models](https://drive.google.com/drive/folders/1ksarh-cJf3F5eKJjLVWY0X1j1qsQqiS2?usp=sharing).
+
+** Update note: Thanks to [Rishikesh (ऋषिकेश)](https://github.com/jaywalnut310/vits/issues/1), our interactive TTS demo is now available on [Colab Notebook](https://colab.research.google.com/drive/1CO61pZizDj7en71NQG_aqqKdGaA_SaBf?usp=sharing).
+
+<table style="width:100%">
+  <tr>
+    <th>VITS at training</th>
+    <th>VITS at inference</th>
+  </tr>
+  <tr>
+    <td><img src="resources/fig_1a.png" alt="VITS at training" height="400"></td>
+    <td><img src="resources/fig_1b.png" alt="VITS at inference" height="400"></td>
+  </tr>
+</table>
+
+
+## Pre-requisites
+0. Python >= 3.6
+0. Clone this repository
+0. Install python requirements. Please refer [requirements.txt](requirements.txt)
+    1. You may need to install espeak first: `apt-get install espeak`
+0. Download datasets
+    1. Download and extract the LJ Speech dataset, then rename or create a link to the dataset folder: `ln -s /path/to/LJSpeech-1.1/wavs DUMMY1`
+    1. For mult-speaker setting, download and extract the VCTK dataset, and downsample wav files to 22050 Hz. Then rename or create a link to the dataset folder: `ln -s /path/to/VCTK-Corpus/downsampled_wavs DUMMY2`
+0. Build Monotonic Alignment Search and run preprocessing if you use your own datasets.
+```sh
+# Cython-version Monotonoic Alignment Search
+cd monotonic_align
+python setup.py build_ext --inplace
+
+# Preprocessing (g2p) for your own datasets. Preprocessed phonemes for LJ Speech and VCTK have been already provided.
+# python preprocess.py --text_index 1 --filelists filelists/ljs_audio_text_train_filelist.txt filelists/ljs_audio_text_val_filelist.txt filelists/ljs_audio_text_test_filelist.txt 
+# python preprocess.py --text_index 2 --filelists filelists/vctk_audio_sid_text_train_filelist.txt filelists/vctk_audio_sid_text_val_filelist.txt filelists/vctk_audio_sid_text_test_filelist.txt
+```
+
+
+## Training Exmaple
+```sh
+# LJ Speech
+python train.py -c configs/ljs_base.json -m ljs_base
+
+# VCTK
+python train_ms.py -c configs/vctk_base.json -m vctk_base
+```
+
+
+## Inference Example
+See [inference.ipynb](inference.ipynb)

attentions.py

@@ -0,0 +1,303 @@
+import copy
+import math
+import numpy as np
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+import commons
+import modules
+from modules import LayerNorm
+   
+
+class Encoder(nn.Module):
+  def __init__(self, hidden_channels, filter_channels, n_heads, n_layers, kernel_size=1, p_dropout=0., window_size=4, **kwargs):
+    super().__init__()
+    self.hidden_channels = hidden_channels
+    self.filter_channels = filter_channels
+    self.n_heads = n_heads
+    self.n_layers = n_layers
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.window_size = window_size
+
+    self.drop = nn.Dropout(p_dropout)
+    self.attn_layers = nn.ModuleList()
+    self.norm_layers_1 = nn.ModuleList()
+    self.ffn_layers = nn.ModuleList()
+    self.norm_layers_2 = nn.ModuleList()
+    for i in range(self.n_layers):
+      self.attn_layers.append(MultiHeadAttention(hidden_channels, hidden_channels, n_heads, p_dropout=p_dropout, window_size=window_size))
+      self.norm_layers_1.append(LayerNorm(hidden_channels))
+      self.ffn_layers.append(FFN(hidden_channels, hidden_channels, filter_channels, kernel_size, p_dropout=p_dropout))
+      self.norm_layers_2.append(LayerNorm(hidden_channels))
+
+  def forward(self, x, x_mask):
+    attn_mask = x_mask.unsqueeze(2) * x_mask.unsqueeze(-1)
+    x = x * x_mask
+    for i in range(self.n_layers):
+      y = self.attn_layers[i](x, x, attn_mask)
+      y = self.drop(y)
+      x = self.norm_layers_1[i](x + y)
+
+      y = self.ffn_layers[i](x, x_mask)
+      y = self.drop(y)
+      x = self.norm_layers_2[i](x + y)
+    x = x * x_mask
+    return x
+
+
+class Decoder(nn.Module):
+  def __init__(self, hidden_channels, filter_channels, n_heads, n_layers, kernel_size=1, p_dropout=0., proximal_bias=False, proximal_init=True, **kwargs):
+    super().__init__()
+    self.hidden_channels = hidden_channels
+    self.filter_channels = filter_channels
+    self.n_heads = n_heads
+    self.n_layers = n_layers
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.proximal_bias = proximal_bias
+    self.proximal_init = proximal_init
+
+    self.drop = nn.Dropout(p_dropout)
+    self.self_attn_layers = nn.ModuleList()
+    self.norm_layers_0 = nn.ModuleList()
+    self.encdec_attn_layers = nn.ModuleList()
+    self.norm_layers_1 = nn.ModuleList()
+    self.ffn_layers = nn.ModuleList()
+    self.norm_layers_2 = nn.ModuleList()
+    for i in range(self.n_layers):
+      self.self_attn_layers.append(MultiHeadAttention(hidden_channels, hidden_channels, n_heads, p_dropout=p_dropout, proximal_bias=proximal_bias, proximal_init=proximal_init))
+      self.norm_layers_0.append(LayerNorm(hidden_channels))
+      self.encdec_attn_layers.append(MultiHeadAttention(hidden_channels, hidden_channels, n_heads, p_dropout=p_dropout))
+      self.norm_layers_1.append(LayerNorm(hidden_channels))
+      self.ffn_layers.append(FFN(hidden_channels, hidden_channels, filter_channels, kernel_size, p_dropout=p_dropout, causal=True))
+      self.norm_layers_2.append(LayerNorm(hidden_channels))
+
+  def forward(self, x, x_mask, h, h_mask):
+    """
+    x: decoder input
+    h: encoder output
+    """
+    self_attn_mask = commons.subsequent_mask(x_mask.size(2)).to(device=x.device, dtype=x.dtype)
+    encdec_attn_mask = h_mask.unsqueeze(2) * x_mask.unsqueeze(-1)
+    x = x * x_mask
+    for i in range(self.n_layers):
+      y = self.self_attn_layers[i](x, x, self_attn_mask)
+      y = self.drop(y)
+      x = self.norm_layers_0[i](x + y)
+
+      y = self.encdec_attn_layers[i](x, h, encdec_attn_mask)
+      y = self.drop(y)
+      x = self.norm_layers_1[i](x + y)
+      
+      y = self.ffn_layers[i](x, x_mask)
+      y = self.drop(y)
+      x = self.norm_layers_2[i](x + y)
+    x = x * x_mask
+    return x
+
+
+class MultiHeadAttention(nn.Module):
+  def __init__(self, channels, out_channels, n_heads, p_dropout=0., window_size=None, heads_share=True, block_length=None, proximal_bias=False, proximal_init=False):
+    super().__init__()
+    assert channels % n_heads == 0
+
+    self.channels = channels
+    self.out_channels = out_channels
+    self.n_heads = n_heads
+    self.p_dropout = p_dropout
+    self.window_size = window_size
+    self.heads_share = heads_share
+    self.block_length = block_length
+    self.proximal_bias = proximal_bias
+    self.proximal_init = proximal_init
+    self.attn = None
+
+    self.k_channels = channels // n_heads
+    self.conv_q = nn.Conv1d(channels, channels, 1)
+    self.conv_k = nn.Conv1d(channels, channels, 1)
+    self.conv_v = nn.Conv1d(channels, channels, 1)
+    self.conv_o = nn.Conv1d(channels, out_channels, 1)
+    self.drop = nn.Dropout(p_dropout)
+
+    if window_size is not None:
+      n_heads_rel = 1 if heads_share else n_heads
+      rel_stddev = self.k_channels**-0.5
+      self.emb_rel_k = nn.Parameter(torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels) * rel_stddev)
+      self.emb_rel_v = nn.Parameter(torch.randn(n_heads_rel, window_size * 2 + 1, self.k_channels) * rel_stddev)
+
+    nn.init.xavier_uniform_(self.conv_q.weight)
+    nn.init.xavier_uniform_(self.conv_k.weight)
+    nn.init.xavier_uniform_(self.conv_v.weight)
+    if proximal_init:
+      with torch.no_grad():
+        self.conv_k.weight.copy_(self.conv_q.weight)
+        self.conv_k.bias.copy_(self.conv_q.bias)
+      
+  def forward(self, x, c, attn_mask=None):
+    q = self.conv_q(x)
+    k = self.conv_k(c)
+    v = self.conv_v(c)
+    
+    x, self.attn = self.attention(q, k, v, mask=attn_mask)
+
+    x = self.conv_o(x)
+    return x
+
+  def attention(self, query, key, value, mask=None):
+    # reshape [b, d, t] -> [b, n_h, t, d_k]
+    b, d, t_s, t_t = (*key.size(), query.size(2))
+    query = query.view(b, self.n_heads, self.k_channels, t_t).transpose(2, 3)
+    key = key.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)
+    value = value.view(b, self.n_heads, self.k_channels, t_s).transpose(2, 3)
+
+    scores = torch.matmul(query / math.sqrt(self.k_channels), key.transpose(-2, -1))
+    if self.window_size is not None:
+      assert t_s == t_t, "Relative attention is only available for self-attention."
+      key_relative_embeddings = self._get_relative_embeddings(self.emb_rel_k, t_s)
+      rel_logits = self._matmul_with_relative_keys(query /math.sqrt(self.k_channels), key_relative_embeddings)
+      scores_local = self._relative_position_to_absolute_position(rel_logits)
+      scores = scores + scores_local
+    if self.proximal_bias:
+      assert t_s == t_t, "Proximal bias is only available for self-attention."
+      scores = scores + self._attention_bias_proximal(t_s).to(device=scores.device, dtype=scores.dtype)
+    if mask is not None:
+      scores = scores.masked_fill(mask == 0, -1e4)
+      if self.block_length is not None:
+        assert t_s == t_t, "Local attention is only available for self-attention."
+        block_mask = torch.ones_like(scores).triu(-self.block_length).tril(self.block_length)
+        scores = scores.masked_fill(block_mask == 0, -1e4)
+    p_attn = F.softmax(scores, dim=-1) # [b, n_h, t_t, t_s]
+    p_attn = self.drop(p_attn)
+    output = torch.matmul(p_attn, value)
+    if self.window_size is not None:
+      relative_weights = self._absolute_position_to_relative_position(p_attn)
+      value_relative_embeddings = self._get_relative_embeddings(self.emb_rel_v, t_s)
+      output = output + self._matmul_with_relative_values(relative_weights, value_relative_embeddings)
+    output = output.transpose(2, 3).contiguous().view(b, d, t_t) # [b, n_h, t_t, d_k] -> [b, d, t_t]
+    return output, p_attn
+
+  def _matmul_with_relative_values(self, x, y):
+    """
+    x: [b, h, l, m]
+    y: [h or 1, m, d]
+    ret: [b, h, l, d]
+    """
+    ret = torch.matmul(x, y.unsqueeze(0))
+    return ret
+
+  def _matmul_with_relative_keys(self, x, y):
+    """
+    x: [b, h, l, d]
+    y: [h or 1, m, d]
+    ret: [b, h, l, m]
+    """
+    ret = torch.matmul(x, y.unsqueeze(0).transpose(-2, -1))
+    return ret
+
+  def _get_relative_embeddings(self, relative_embeddings, length):
+    max_relative_position = 2 * self.window_size + 1
+    # Pad first before slice to avoid using cond ops.
+    pad_length = max(length - (self.window_size + 1), 0)
+    slice_start_position = max((self.window_size + 1) - length, 0)
+    slice_end_position = slice_start_position + 2 * length - 1
+    if pad_length > 0:
+      padded_relative_embeddings = F.pad(
+          relative_embeddings,
+          commons.convert_pad_shape([[0, 0], [pad_length, pad_length], [0, 0]]))
+    else:
+      padded_relative_embeddings = relative_embeddings
+    used_relative_embeddings = padded_relative_embeddings[:,slice_start_position:slice_end_position]
+    return used_relative_embeddings
+
+  def _relative_position_to_absolute_position(self, x):
+    """
+    x: [b, h, l, 2*l-1]
+    ret: [b, h, l, l]
+    """
+    batch, heads, length, _ = x.size()
+    # Concat columns of pad to shift from relative to absolute indexing.
+    x = F.pad(x, commons.convert_pad_shape([[0,0],[0,0],[0,0],[0,1]]))
+
+    # Concat extra elements so to add up to shape (len+1, 2*len-1).
+    x_flat = x.view([batch, heads, length * 2 * length])
+    x_flat = F.pad(x_flat, commons.convert_pad_shape([[0,0],[0,0],[0,length-1]]))
+
+    # Reshape and slice out the padded elements.
+    x_final = x_flat.view([batch, heads, length+1, 2*length-1])[:, :, :length, length-1:]
+    return x_final
+
+  def _absolute_position_to_relative_position(self, x):
+    """
+    x: [b, h, l, l]
+    ret: [b, h, l, 2*l-1]
+    """
+    batch, heads, length, _ = x.size()
+    # padd along column
+    x = F.pad(x, commons.convert_pad_shape([[0, 0], [0, 0], [0, 0], [0, length-1]]))
+    x_flat = x.view([batch, heads, length**2 + length*(length -1)])
+    # add 0's in the beginning that will skew the elements after reshape
+    x_flat = F.pad(x_flat, commons.convert_pad_shape([[0, 0], [0, 0], [length, 0]]))
+    x_final = x_flat.view([batch, heads, length, 2*length])[:,:,:,1:]
+    return x_final
+
+  def _attention_bias_proximal(self, length):
+    """Bias for self-attention to encourage attention to close positions.
+    Args:
+      length: an integer scalar.
+    Returns:
+      a Tensor with shape [1, 1, length, length]
+    """
+    r = torch.arange(length, dtype=torch.float32)
+    diff = torch.unsqueeze(r, 0) - torch.unsqueeze(r, 1)
+    return torch.unsqueeze(torch.unsqueeze(-torch.log1p(torch.abs(diff)), 0), 0)
+
+
+class FFN(nn.Module):
+  def __init__(self, in_channels, out_channels, filter_channels, kernel_size, p_dropout=0., activation=None, causal=False):
+    super().__init__()
+    self.in_channels = in_channels
+    self.out_channels = out_channels
+    self.filter_channels = filter_channels
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.activation = activation
+    self.causal = causal
+
+    if causal:
+      self.padding = self._causal_padding
+    else:
+      self.padding = self._same_padding
+
+    self.conv_1 = nn.Conv1d(in_channels, filter_channels, kernel_size)
+    self.conv_2 = nn.Conv1d(filter_channels, out_channels, kernel_size)
+    self.drop = nn.Dropout(p_dropout)
+
+  def forward(self, x, x_mask):
+    x = self.conv_1(self.padding(x * x_mask))
+    if self.activation == "gelu":
+      x = x * torch.sigmoid(1.702 * x)
+    else:
+      x = torch.relu(x)
+    x = self.drop(x)
+    x = self.conv_2(self.padding(x * x_mask))
+    return x * x_mask
+  
+  def _causal_padding(self, x):
+    if self.kernel_size == 1:
+      return x
+    pad_l = self.kernel_size - 1
+    pad_r = 0
+    padding = [[0, 0], [0, 0], [pad_l, pad_r]]
+    x = F.pad(x, commons.convert_pad_shape(padding))
+    return x
+
+  def _same_padding(self, x):
+    if self.kernel_size == 1:
+      return x
+    pad_l = (self.kernel_size - 1) // 2
+    pad_r = self.kernel_size // 2
+    padding = [[0, 0], [0, 0], [pad_l, pad_r]]
+    x = F.pad(x, commons.convert_pad_shape(padding))
+    return x

colab.ipynb

@@ -0,0 +1 @@
+{"cells":[{"cell_type":"markdown","metadata":{"id":"65CU-n-JHhbY"},"source":["# Clone repository"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":1000},"executionInfo":{"elapsed":198272,"status":"ok","timestamp":1659461977037,"user":{"displayName":"章政","userId":"17693849672782836082"},"user_tz":-480},"id":"i_0vZ-OjHVNu","outputId":"52655c4e-699c-465a-ce12-9cef24aa8a1e"},"outputs":[],"source":["!git clone https://github.com/CjangCjengh/vits.git\n","%cd vits\n","!pip install -r requirements.txt\n","!sudo apt-get install espeak -y"]},{"cell_type":"markdown","metadata":{"id":"G0iLn2JxKYhl"},"source":["# Mount Google Drive"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":2801,"status":"ok","timestamp":1659464479547,"user":{"displayName":"章政","userId":"17693849672782836082"},"user_tz":-480},"id":"ZOgjdsQgKTfD","outputId":"ba0ce34b-45f6-43ea-af98-98e6007a5351"},"outputs":[],"source":["from google.colab import drive\n","drive.mount('/content/drive')"]},{"cell_type":"markdown","metadata":{"id":"UZ9maSoUmHaS"},"source":["# Unpack dataset"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":191445,"status":"ok","timestamp":1659462201164,"user":{"displayName":"章政","userId":"17693849672782836082"},"user_tz":-480},"id":"N3a-FsHghwXS","outputId":"979fe3a3-49ba-4f0d-cc71-fc68b8ec0db7"},"outputs":[],"source":["!sudo apt-get install p7zip-full\n","!7z x ../drive/MyDrive/dataset.zip"]},{"cell_type":"markdown","metadata":{"id":"LY9d2hgjmYUF"},"source":["# Alignment"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":4802,"status":"ok","timestamp":1659462298052,"user":{"displayName":"章政","userId":"17693849672782836082"},"user_tz":-480},"id":"LOsV22D8IUTS","outputId":"8b9b7c35-e869-4af5-a8eb-9e6eee11fa7c"},"outputs":[],"source":["%cd monotonic_align\n","!python setup.py build_ext --inplace\n","%cd .."]},{"cell_type":"markdown","metadata":{"id":"gjIAR_UsmPEz"},"source":["# Train"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":245907,"status":"ok","timestamp":1659464394347,"user":{"displayName":"章政","userId":"17693849672782836082"},"user_tz":-480},"id":"ltU2JXpxIh-K","outputId":"13573700-ca5d-4bd1-ebbc-966c059e1327"},"outputs":[],"source":["!python train_ms.py -c configs/yuzusoft_base.json -m yuzusoft_base"]}],"metadata":{"accelerator":"GPU","colab":{"authorship_tag":"ABX9TyMUFamkuoxXK2DqqYNB4cPL","collapsed_sections":[],"mount_file_id":"1twKgqwggarlmTzNEjCsuldastf_zMIkF","name":"vits.ipynb","provenance":[]},"gpuClass":"standard","kernelspec":{"display_name":"Python 3","name":"python3"},"language_info":{"name":"python"}},"nbformat":4,"nbformat_minor":0}

commons.py

@@ -0,0 +1,161 @@
+import math
+import numpy as np
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+
+def init_weights(m, mean=0.0, std=0.01):
+  classname = m.__class__.__name__
+  if classname.find("Conv") != -1:
+    m.weight.data.normal_(mean, std)
+
+
+def get_padding(kernel_size, dilation=1):
+  return int((kernel_size*dilation - dilation)/2)
+
+
+def convert_pad_shape(pad_shape):
+  l = pad_shape[::-1]
+  pad_shape = [item for sublist in l for item in sublist]
+  return pad_shape
+
+
+def intersperse(lst, item):
+  result = [item] * (len(lst) * 2 + 1)
+  result[1::2] = lst
+  return result
+
+
+def kl_divergence(m_p, logs_p, m_q, logs_q):
+  """KL(P||Q)"""
+  kl = (logs_q - logs_p) - 0.5
+  kl += 0.5 * (torch.exp(2. * logs_p) + ((m_p - m_q)**2)) * torch.exp(-2. * logs_q)
+  return kl
+
+
+def rand_gumbel(shape):
+  """Sample from the Gumbel distribution, protect from overflows."""
+  uniform_samples = torch.rand(shape) * 0.99998 + 0.00001
+  return -torch.log(-torch.log(uniform_samples))
+
+
+def rand_gumbel_like(x):
+  g = rand_gumbel(x.size()).to(dtype=x.dtype, device=x.device)
+  return g
+
+
+def slice_segments(x, ids_str, segment_size=4):
+  ret = torch.zeros_like(x[:, :, :segment_size])
+  for i in range(x.size(0)):
+    idx_str = ids_str[i]
+    idx_end = idx_str + segment_size
+    ret[i] = x[i, :, idx_str:idx_end]
+  return ret
+
+
+def rand_slice_segments(x, x_lengths=None, segment_size=4):
+  b, d, t = x.size()
+  if x_lengths is None:
+    x_lengths = t
+  ids_str_max = x_lengths - segment_size + 1
+  ids_str = (torch.rand([b]).to(device=x.device) * ids_str_max).to(dtype=torch.long)
+  ret = slice_segments(x, ids_str, segment_size)
+  return ret, ids_str
+
+
+def get_timing_signal_1d(
+    length, channels, min_timescale=1.0, max_timescale=1.0e4):
+  position = torch.arange(length, dtype=torch.float)
+  num_timescales = channels // 2
+  log_timescale_increment = (
+      math.log(float(max_timescale) / float(min_timescale)) /
+      (num_timescales - 1))
+  inv_timescales = min_timescale * torch.exp(
+      torch.arange(num_timescales, dtype=torch.float) * -log_timescale_increment)
+  scaled_time = position.unsqueeze(0) * inv_timescales.unsqueeze(1)
+  signal = torch.cat([torch.sin(scaled_time), torch.cos(scaled_time)], 0)
+  signal = F.pad(signal, [0, 0, 0, channels % 2])
+  signal = signal.view(1, channels, length)
+  return signal
+
+
+def add_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4):
+  b, channels, length = x.size()
+  signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale)
+  return x + signal.to(dtype=x.dtype, device=x.device)
+
+
+def cat_timing_signal_1d(x, min_timescale=1.0, max_timescale=1.0e4, axis=1):
+  b, channels, length = x.size()
+  signal = get_timing_signal_1d(length, channels, min_timescale, max_timescale)
+  return torch.cat([x, signal.to(dtype=x.dtype, device=x.device)], axis)
+
+
+def subsequent_mask(length):
+  mask = torch.tril(torch.ones(length, length)).unsqueeze(0).unsqueeze(0)
+  return mask
+
+
+@torch.jit.script
+def fused_add_tanh_sigmoid_multiply(input_a, input_b, n_channels):
+  n_channels_int = n_channels[0]
+  in_act = input_a + input_b
+  t_act = torch.tanh(in_act[:, :n_channels_int, :])
+  s_act = torch.sigmoid(in_act[:, n_channels_int:, :])
+  acts = t_act * s_act
+  return acts
+
+
+def convert_pad_shape(pad_shape):
+  l = pad_shape[::-1]
+  pad_shape = [item for sublist in l for item in sublist]
+  return pad_shape
+
+
+def shift_1d(x):
+  x = F.pad(x, convert_pad_shape([[0, 0], [0, 0], [1, 0]]))[:, :, :-1]
+  return x
+
+
+def sequence_mask(length, max_length=None):
+  if max_length is None:
+    max_length = length.max()
+  x = torch.arange(max_length, dtype=length.dtype, device=length.device)
+  return x.unsqueeze(0) < length.unsqueeze(1)
+
+
+def generate_path(duration, mask):
+  """
+  duration: [b, 1, t_x]
+  mask: [b, 1, t_y, t_x]
+  """
+  device = duration.device
+  
+  b, _, t_y, t_x = mask.shape
+  cum_duration = torch.cumsum(duration, -1)
+  
+  cum_duration_flat = cum_duration.view(b * t_x)
+  path = sequence_mask(cum_duration_flat, t_y).to(mask.dtype)
+  path = path.view(b, t_x, t_y)
+  path = path - F.pad(path, convert_pad_shape([[0, 0], [1, 0], [0, 0]]))[:, :-1]
+  path = path.unsqueeze(1).transpose(2,3) * mask
+  return path
+
+
+def clip_grad_value_(parameters, clip_value, norm_type=2):
+  if isinstance(parameters, torch.Tensor):
+    parameters = [parameters]
+  parameters = list(filter(lambda p: p.grad is not None, parameters))
+  norm_type = float(norm_type)
+  if clip_value is not None:
+    clip_value = float(clip_value)
+
+  total_norm = 0
+  for p in parameters:
+    param_norm = p.grad.data.norm(norm_type)
+    total_norm += param_norm.item() ** norm_type
+    if clip_value is not None:
+      p.grad.data.clamp_(min=-clip_value, max=clip_value)
+  total_norm = total_norm ** (1. / norm_type)
+  return total_norm

configs/japanese_base.json

@@ -0,0 +1,55 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 2000,
+    "seed": 1234,
+    "epochs": 10000,
+    "learning_rate": 2e-4,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 24,
+    "fp16_run": true,
+    "lr_decay": 0.999875,
+    "segment_size": 8192,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0
+  },
+  "data": {
+    "training_files":"filelists/trainnn.txt.cleaned",
+    "validation_files":"filelists/val.txt.cleaned",
+    "text_cleaners":["japanese_cleaners"],
+    "max_wav_value": 32768.0,
+    "sampling_rate": 22050,
+    "filter_length": 1024,
+    "hop_length": 256,
+    "win_length": 1024,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 7,
+    "cleaned_text": true
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 192,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 6,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [3,7,11],
+    "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
+    "upsample_rates": [8,8,2,2],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [16,16,4,4],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 256
+  },
+  "speakers": ["\u7dbe\u5730\u5be7\u3005", "\u56e0\u5e61\u3081\u3050\u308b", "\u671d\u6b66\u82b3\u4e43", "\u5e38\u9678\u8309\u5b50", "\u30e0\u30e9\u30b5\u30e1", "\u978d\u99ac\u5c0f\u6625", "\u5728\u539f\u4e03\u6d77"],
+  "symbols": ["_", ",", ".", "!", "?", "-", "A", "E", "I", "N", "O", "Q", "U", "a", "b", "d", "e", "f", "g", "h", "i", "j", "k", "m", "n", "o", "p", "r", "s", "t", "u", "v", "w", "y", "z", "\u0283", "\u02a7", "\u2193", "\u2191", " "]
+}

configs/japanese_base2.json

@@ -0,0 +1,55 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 1000,
+    "seed": 1234,
+    "epochs": 10000,
+    "learning_rate": 2e-4,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 32,
+    "fp16_run": true,
+    "lr_decay": 0.999875,
+    "segment_size": 8192,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0
+  },
+  "data": {
+    "training_files":"filelists/hamidashi_train_filelist.txt.cleaned",
+    "validation_files":"filelists/hamidashi_val_filelist.txt.cleaned",
+    "text_cleaners":["japanese_cleaners2"],
+    "max_wav_value": 32768.0,
+    "sampling_rate": 22050,
+    "filter_length": 1024,
+    "hop_length": 256,
+    "win_length": 1024,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 8,
+    "cleaned_text": true
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 192,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 6,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [3,7,11],
+    "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
+    "upsample_rates": [8,8,2,2],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [16,16,4,4],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 256
+  },
+  "speakers": ["\u548c\u6cc9\u5983\u611b", "\u5e38\u76e4\u83ef\u4e43", "\u9326\u3042\u3059\u307f", "\u938c\u5009\u8a69\u685c", "\u7adc\u9591\u5929\u68a8", "\u548c\u6cc9\u91cc", "\u65b0\u5ddd\u5e83\u5922", "\u8056\u8389\u3005\u5b50"],
+  "symbols": ["_", ",", ".", "!", "?", "-", "~", "\u2026", "A", "E", "I", "N", "O", "Q", "U", "a", "b", "d", "e", "f", "g", "h", "i", "j", "k", "m", "n", "o", "p", "r", "s", "t", "u", "v", "w", "y", "z", "\u0283", "\u02a7", "\u02a6", "\u2193", "\u2191", " "]
+}

configs/japanese_ss_base2.json

@@ -0,0 +1,54 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 1000,
+    "seed": 1234,
+    "epochs": 20000,
+    "learning_rate": 2e-4,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 32,
+    "fp16_run": true,
+    "lr_decay": 0.999875,
+    "segment_size": 8192,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0
+  },
+  "data": {
+    "training_files":"filelists/train_filelist.txt.cleaned",
+    "validation_files":"filelists/val_filelist.txt.cleaned",
+    "text_cleaners":["japanese_cleaners2"],
+    "max_wav_value": 32768.0,
+    "sampling_rate": 22050,
+    "filter_length": 1024,
+    "hop_length": 256,
+    "win_length": 1024,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 0,
+    "cleaned_text": true
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 192,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 6,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [3,7,11],
+    "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
+    "upsample_rates": [8,8,2,2],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [16,16,4,4],
+    "n_layers_q": 3,
+    "use_spectral_norm": false
+  },
+  "speakers": ["\u30eb\u30a4\u30ba"],
+  "symbols": ["_", ",", ".", "!", "?", "-", "~", "\u2026", "A", "E", "I", "N", "O", "Q", "U", "a", "b", "d", "e", "f", "g", "h", "i", "j", "k", "m", "n", "o", "p", "r", "s", "t", "u", "v", "w", "y", "z", "\u0283", "\u02a7", "\u02a6", "\u2193", "\u2191", " "]
+}

configs/korean_base.json

@@ -0,0 +1,55 @@
+{
+  "train": {
+    "log_interval": 200,
+    "eval_interval": 1000,
+    "seed": 1234,
+    "epochs": 10000,
+    "learning_rate": 2e-4,
+    "betas": [0.8, 0.99],
+    "eps": 1e-9,
+    "batch_size": 32,
+    "fp16_run": true,
+    "lr_decay": 0.999875,
+    "segment_size": 8192,
+    "init_lr_ratio": 1,
+    "warmup_epochs": 0,
+    "c_mel": 45,
+    "c_kl": 1.0
+  },
+  "data": {
+    "training_files":"filelists/train_filelist.txt.cleaned",
+    "validation_files":"filelists/val_filelist.txt.cleaned",
+    "text_cleaners":["korean_cleaners"],
+    "max_wav_value": 32768.0,
+    "sampling_rate": 22050,
+    "filter_length": 1024,
+    "hop_length": 256,
+    "win_length": 1024,
+    "n_mel_channels": 80,
+    "mel_fmin": 0.0,
+    "mel_fmax": null,
+    "add_blank": true,
+    "n_speakers": 6,
+    "cleaned_text": true
+  },
+  "model": {
+    "inter_channels": 192,
+    "hidden_channels": 192,
+    "filter_channels": 768,
+    "n_heads": 2,
+    "n_layers": 6,
+    "kernel_size": 3,
+    "p_dropout": 0.1,
+    "resblock": "1",
+    "resblock_kernel_sizes": [3,7,11],
+    "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
+    "upsample_rates": [8,8,2,2],
+    "upsample_initial_channel": 512,
+    "upsample_kernel_sizes": [16,16,4,4],
+    "n_layers_q": 3,
+    "use_spectral_norm": false,
+    "gin_channels": 256
+  },
+  "speakers": ["\uc218\uc544", "\ubbf8\ubbf8\ub974", "\uc544\ub9b0", "\uc5f0\ud654", "\uc720\ud654", "\uc120\ubc30"],
+  "symbols": ["_", ",", ".", "!", "?", "\u2026", "~", "\u3131", "\u3134", "\u3137", "\u3139", "\u3141", "\u3142", "\u3145", "\u3147", "\u3148", "\u314a", "\u314b", "\u314c", "\u314d", "\u314e", "\u3132", "\u3138", "\u3143", "\u3146", "\u3149", "\u314f", "\u3153", "\u3157", "\u315c", "\u3161", "\u3163", "\u3150", "\u3154", " "]
+}

data_utils.py

@@ -0,0 +1,392 @@
+import time
+import os
+import random
+import numpy as np
+import torch
+import torch.utils.data
+
+import commons 
+from mel_processing import spectrogram_torch
+from utils import load_wav_to_torch, load_filepaths_and_text
+from text import text_to_sequence, cleaned_text_to_sequence
+
+
+class TextAudioLoader(torch.utils.data.Dataset):
+    """
+        1) loads audio, text pairs
+        2) normalizes text and converts them to sequences of integers
+        3) computes spectrograms from audio files.
+    """
+    def __init__(self, audiopaths_and_text, hparams):
+        self.audiopaths_and_text = load_filepaths_and_text(audiopaths_and_text)
+        self.text_cleaners  = hparams.text_cleaners
+        self.max_wav_value  = hparams.max_wav_value
+        self.sampling_rate  = hparams.sampling_rate
+        self.filter_length  = hparams.filter_length 
+        self.hop_length     = hparams.hop_length 
+        self.win_length     = hparams.win_length
+        self.sampling_rate  = hparams.sampling_rate 
+
+        self.cleaned_text = getattr(hparams, "cleaned_text", False)
+
+        self.add_blank = hparams.add_blank
+        self.min_text_len = getattr(hparams, "min_text_len", 1)
+        self.max_text_len = getattr(hparams, "max_text_len", 190)
+
+        random.seed(1234)
+        random.shuffle(self.audiopaths_and_text)
+        self._filter()
+
+
+    def _filter(self):
+        """
+        Filter text & store spec lengths
+        """
+        # Store spectrogram lengths for Bucketing
+        # wav_length ~= file_size / (wav_channels * Bytes per dim) = file_size / (1 * 2)
+        # spec_length = wav_length // hop_length
+
+        audiopaths_and_text_new = []
+        lengths = []
+        for audiopath, text in self.audiopaths_and_text:
+            if self.min_text_len <= len(text) and len(text) <= self.max_text_len:
+                audiopaths_and_text_new.append([audiopath, text])
+                lengths.append(os.path.getsize(audiopath) // (2 * self.hop_length))
+        self.audiopaths_and_text = audiopaths_and_text_new
+        self.lengths = lengths
+
+    def get_audio_text_pair(self, audiopath_and_text):
+        # separate filename and text
+        audiopath, text = audiopath_and_text[0], audiopath_and_text[1]
+        text = self.get_text(text)
+        spec, wav = self.get_audio(audiopath)
+        return (text, spec, wav)
+
+    def get_audio(self, filename):
+        audio, sampling_rate = load_wav_to_torch(filename)
+        if sampling_rate != self.sampling_rate:
+            raise ValueError("{} {} SR doesn't match target {} SR".format(
+                sampling_rate, self.sampling_rate))
+        audio_norm = audio / self.max_wav_value
+        audio_norm = audio_norm.unsqueeze(0)
+        spec_filename = filename.replace(".wav", ".spec.pt")
+        if os.path.exists(spec_filename):
+            spec = torch.load(spec_filename)
+        else:
+            spec = spectrogram_torch(audio_norm, self.filter_length,
+                self.sampling_rate, self.hop_length, self.win_length,
+                center=False)
+            spec = torch.squeeze(spec, 0)
+            torch.save(spec, spec_filename)
+        return spec, audio_norm
+
+    def get_text(self, text):
+        if self.cleaned_text:
+            text_norm = cleaned_text_to_sequence(text)
+        else:
+            text_norm = text_to_sequence(text, self.text_cleaners)
+        if self.add_blank:
+            text_norm = commons.intersperse(text_norm, 0)
+        text_norm = torch.LongTensor(text_norm)
+        return text_norm
+
+    def __getitem__(self, index):
+        return self.get_audio_text_pair(self.audiopaths_and_text[index])
+
+    def __len__(self):
+        return len(self.audiopaths_and_text)
+
+
+class TextAudioCollate():
+    """ Zero-pads model inputs and targets
+    """
+    def __init__(self, return_ids=False):
+        self.return_ids = return_ids
+
+    def __call__(self, batch):
+        """Collate's training batch from normalized text and aduio
+        PARAMS
+        ------
+        batch: [text_normalized, spec_normalized, wav_normalized]
+        """
+        # Right zero-pad all one-hot text sequences to max input length
+        _, ids_sorted_decreasing = torch.sort(
+            torch.LongTensor([x[1].size(1) for x in batch]),
+            dim=0, descending=True)
+
+        max_text_len = max([len(x[0]) for x in batch])
+        max_spec_len = max([x[1].size(1) for x in batch])
+        max_wav_len = max([x[2].size(1) for x in batch])
+
+        text_lengths = torch.LongTensor(len(batch))
+        spec_lengths = torch.LongTensor(len(batch))
+        wav_lengths = torch.LongTensor(len(batch))
+
+        text_padded = torch.LongTensor(len(batch), max_text_len)
+        spec_padded = torch.FloatTensor(len(batch), batch[0][1].size(0), max_spec_len)
+        wav_padded = torch.FloatTensor(len(batch), 1, max_wav_len)
+        text_padded.zero_()
+        spec_padded.zero_()
+        wav_padded.zero_()
+        for i in range(len(ids_sorted_decreasing)):
+            row = batch[ids_sorted_decreasing[i]]
+
+            text = row[0]
+            text_padded[i, :text.size(0)] = text
+            text_lengths[i] = text.size(0)
+
+            spec = row[1]
+            spec_padded[i, :, :spec.size(1)] = spec
+            spec_lengths[i] = spec.size(1)
+
+            wav = row[2]
+            wav_padded[i, :, :wav.size(1)] = wav
+            wav_lengths[i] = wav.size(1)
+
+        if self.return_ids:
+            return text_padded, text_lengths, spec_padded, spec_lengths, wav_padded, wav_lengths, ids_sorted_decreasing
+        return text_padded, text_lengths, spec_padded, spec_lengths, wav_padded, wav_lengths
+
+
+"""Multi speaker version"""
+class TextAudioSpeakerLoader(torch.utils.data.Dataset):
+    """
+        1) loads audio, speaker_id, text pairs
+        2) normalizes text and converts them to sequences of integers
+        3) computes spectrograms from audio files.
+    """
+    def __init__(self, audiopaths_sid_text, hparams):
+        self.audiopaths_sid_text = load_filepaths_and_text(audiopaths_sid_text)
+        self.text_cleaners = hparams.text_cleaners
+        self.max_wav_value = hparams.max_wav_value
+        self.sampling_rate = hparams.sampling_rate
+        self.filter_length  = hparams.filter_length
+        self.hop_length     = hparams.hop_length
+        self.win_length     = hparams.win_length
+        self.sampling_rate  = hparams.sampling_rate
+
+        self.cleaned_text = getattr(hparams, "cleaned_text", False)
+
+        self.add_blank = hparams.add_blank
+        self.min_text_len = getattr(hparams, "min_text_len", 1)
+        self.max_text_len = getattr(hparams, "max_text_len", 190)
+
+        random.seed(1234)
+        random.shuffle(self.audiopaths_sid_text)
+        self._filter()
+
+    def _filter(self):
+        """
+        Filter text & store spec lengths
+        """
+        # Store spectrogram lengths for Bucketing
+        # wav_length ~= file_size / (wav_channels * Bytes per dim) = file_size / (1 * 2)
+        # spec_length = wav_length // hop_length
+
+        audiopaths_sid_text_new = []
+        lengths = []
+        for audiopath, sid, text in self.audiopaths_sid_text:
+            if self.min_text_len <= len(text) and len(text) <= self.max_text_len:
+                audiopaths_sid_text_new.append([audiopath, sid, text])
+                lengths.append(os.path.getsize(audiopath) // (2 * self.hop_length))
+        self.audiopaths_sid_text = audiopaths_sid_text_new
+        self.lengths = lengths
+
+    def get_audio_text_speaker_pair(self, audiopath_sid_text):
+        # separate filename, speaker_id and text
+        audiopath, sid, text = audiopath_sid_text[0], audiopath_sid_text[1], audiopath_sid_text[2]
+        text = self.get_text(text)
+        spec, wav = self.get_audio(audiopath)
+        sid = self.get_sid(sid)
+        return (text, spec, wav, sid)
+
+    def get_audio(self, filename):
+        audio, sampling_rate = load_wav_to_torch(filename)
+        if sampling_rate != self.sampling_rate:
+            raise ValueError("{} {} SR doesn't match target {} SR".format(
+                sampling_rate, self.sampling_rate))
+        audio_norm = audio / self.max_wav_value
+        audio_norm = audio_norm.unsqueeze(0)
+        spec_filename = filename.replace(".wav", ".spec.pt")
+        if os.path.exists(spec_filename):
+            spec = torch.load(spec_filename)
+        else:
+            spec = spectrogram_torch(audio_norm, self.filter_length,
+                self.sampling_rate, self.hop_length, self.win_length,
+                center=False)
+            spec = torch.squeeze(spec, 0)
+            torch.save(spec, spec_filename)
+        return spec, audio_norm
+
+    def get_text(self, text):
+        if self.cleaned_text:
+            text_norm = cleaned_text_to_sequence(text)
+        else:
+            text_norm = text_to_sequence(text, self.text_cleaners)
+        if self.add_blank:
+            text_norm = commons.intersperse(text_norm, 0)
+        text_norm = torch.LongTensor(text_norm)
+        return text_norm
+
+    def get_sid(self, sid):
+        sid = torch.LongTensor([int(sid)])
+        return sid
+
+    def __getitem__(self, index):
+        return self.get_audio_text_speaker_pair(self.audiopaths_sid_text[index])
+
+    def __len__(self):
+        return len(self.audiopaths_sid_text)
+
+
+class TextAudioSpeakerCollate():
+    """ Zero-pads model inputs and targets
+    """
+    def __init__(self, return_ids=False):
+        self.return_ids = return_ids
+
+    def __call__(self, batch):
+        """Collate's training batch from normalized text, audio and speaker identities
+        PARAMS
+        ------
+        batch: [text_normalized, spec_normalized, wav_normalized, sid]
+        """
+        # Right zero-pad all one-hot text sequences to max input length
+        _, ids_sorted_decreasing = torch.sort(
+            torch.LongTensor([x[1].size(1) for x in batch]),
+            dim=0, descending=True)
+
+        max_text_len = max([len(x[0]) for x in batch])
+        max_spec_len = max([x[1].size(1) for x in batch])
+        max_wav_len = max([x[2].size(1) for x in batch])
+
+        text_lengths = torch.LongTensor(len(batch))
+        spec_lengths = torch.LongTensor(len(batch))
+        wav_lengths = torch.LongTensor(len(batch))
+        sid = torch.LongTensor(len(batch))
+
+        text_padded = torch.LongTensor(len(batch), max_text_len)
+        spec_padded = torch.FloatTensor(len(batch), batch[0][1].size(0), max_spec_len)
+        wav_padded = torch.FloatTensor(len(batch), 1, max_wav_len)
+        text_padded.zero_()
+        spec_padded.zero_()
+        wav_padded.zero_()
+        for i in range(len(ids_sorted_decreasing)):
+            row = batch[ids_sorted_decreasing[i]]
+
+            text = row[0]
+            text_padded[i, :text.size(0)] = text
+            text_lengths[i] = text.size(0)
+
+            spec = row[1]
+            spec_padded[i, :, :spec.size(1)] = spec
+            spec_lengths[i] = spec.size(1)
+
+            wav = row[2]
+            wav_padded[i, :, :wav.size(1)] = wav
+            wav_lengths[i] = wav.size(1)
+
+            sid[i] = row[3]
+
+        if self.return_ids:
+            return text_padded, text_lengths, spec_padded, spec_lengths, wav_padded, wav_lengths, sid, ids_sorted_decreasing
+        return text_padded, text_lengths, spec_padded, spec_lengths, wav_padded, wav_lengths, sid
+
+
+class DistributedBucketSampler(torch.utils.data.distributed.DistributedSampler):
+    """
+    Maintain similar input lengths in a batch.
+    Length groups are specified by boundaries.
+    Ex) boundaries = [b1, b2, b3] -> any batch is included either {x | b1 < length(x) <=b2} or {x | b2 < length(x) <= b3}.
+  
+    It removes samples which are not included in the boundaries.
+    Ex) boundaries = [b1, b2, b3] -> any x s.t. length(x) <= b1 or length(x) > b3 are discarded.
+    """
+    def __init__(self, dataset, batch_size, boundaries, num_replicas=None, rank=None, shuffle=True):
+        super().__init__(dataset, num_replicas=num_replicas, rank=rank, shuffle=shuffle)
+        self.lengths = dataset.lengths
+        self.batch_size = batch_size
+        self.boundaries = boundaries
+  
+        self.buckets, self.num_samples_per_bucket = self._create_buckets()
+        self.total_size = sum(self.num_samples_per_bucket)
+        self.num_samples = self.total_size // self.num_replicas
+  
+    def _create_buckets(self):
+        buckets = [[] for _ in range(len(self.boundaries) - 1)]
+        for i in range(len(self.lengths)):
+            length = self.lengths[i]
+            idx_bucket = self._bisect(length)
+            if idx_bucket != -1:
+                buckets[idx_bucket].append(i)
+  
+        for i in range(len(buckets) - 1, 0, -1):
+            if len(buckets[i]) == 0:
+                buckets.pop(i)
+                self.boundaries.pop(i+1)
+  
+        num_samples_per_bucket = []
+        for i in range(len(buckets)):
+            len_bucket = len(buckets[i])
+            total_batch_size = self.num_replicas * self.batch_size
+            rem = (total_batch_size - (len_bucket % total_batch_size)) % total_batch_size
+            num_samples_per_bucket.append(len_bucket + rem)
+        return buckets, num_samples_per_bucket
+  
+    def __iter__(self):
+      # deterministically shuffle based on epoch
+      g = torch.Generator()
+      g.manual_seed(self.epoch)
+  
+      indices = []
+      if self.shuffle:
+          for bucket in self.buckets:
+              indices.append(torch.randperm(len(bucket), generator=g).tolist())
+      else:
+          for bucket in self.buckets:
+              indices.append(list(range(len(bucket))))
+  
+      batches = []
+      for i in range(len(self.buckets)):
+          bucket = self.buckets[i]
+          len_bucket = len(bucket)
+          ids_bucket = indices[i]
+          num_samples_bucket = self.num_samples_per_bucket[i]
+  
+          # add extra samples to make it evenly divisible
+          rem = num_samples_bucket - len_bucket
+          ids_bucket = ids_bucket + ids_bucket * (rem // len_bucket) + ids_bucket[:(rem % len_bucket)]
+  
+          # subsample
+          ids_bucket = ids_bucket[self.rank::self.num_replicas]
+  
+          # batching
+          for j in range(len(ids_bucket) // self.batch_size):
+              batch = [bucket[idx] for idx in ids_bucket[j*self.batch_size:(j+1)*self.batch_size]]
+              batches.append(batch)
+  
+      if self.shuffle:
+          batch_ids = torch.randperm(len(batches), generator=g).tolist()
+          batches = [batches[i] for i in batch_ids]
+      self.batches = batches
+  
+      assert len(self.batches) * self.batch_size == self.num_samples
+      return iter(self.batches)
+  
+    def _bisect(self, x, lo=0, hi=None):
+      if hi is None:
+          hi = len(self.boundaries) - 1
+  
+      if hi > lo:
+          mid = (hi + lo) // 2
+          if self.boundaries[mid] < x and x <= self.boundaries[mid+1]:
+              return mid
+          elif x <= self.boundaries[mid]:
+              return self._bisect(x, lo, mid)
+          else:
+              return self._bisect(x, mid + 1, hi)
+      else:
+          return -1
+
+    def __len__(self):
+        return self.num_samples // self.batch_size

filelists/hamidashi_val_filelist.txt

@@ -0,0 +1,73 @@
+dataset/hiy/fem_hiy_12064.wav|0|あれあれお兄がいつになくだらしない顔でにこにこしているぞ？　なに見てるのかな？　私の画像？
+dataset/asu/fem_asu_10300.wav|2|は、はい……！　ありがとうございます
+dataset/hiy/fem_hiy_12116.wav|0|大丈夫。夢じゃないよ。このひよりが保証するよ
+dataset/asu/fem_asu_10635.wav|2|せんぱい……？
+dataset/asu/fem_asu_10789.wav|2|不登校の人たちに、出てくるのも悪くないよって伝えるためのものだと思っているので……
+dataset/asu/fem_asu_10912.wav|2|ＶＴｕｂｅｒは一つの可能性なんです。雪景シキは、それを応援したいです
+dataset/kan/fem_kan_12215.wav|1|き、気のせいよ……
+dataset/hir/man_hir_10106.wav|6|代役……雨野葉レマが出るわけにはいきませんか？
+dataset/asu/fem_asu_11205.wav|2|いえ、鳴門さんが女性のかたを一人手配してくれたので、その人と一緒に行ってきます
+dataset/asu/fem_asu_11328.wav|2|でも舞台に立ってわかったんです。応援……声援って、とっても嬉しいんです。ぐんぐん力が湧いてきます
+dataset/asu/fem_asu_11523.wav|2|えっと、それは、その……せんぱいが、気持ちよく、するから……
+dataset/asu/fem_asu_11723.wav|2|ちゅるる、んっ、ちゅ……ちゅぷ、じゅる、れる……じゅる……じゅぷ……じゅるる、ん……じゅぷぷ、んぅ……じゅぷ、じゅぷぷぷ……
+dataset/asu/fem_asu_11926.wav|2|ごめんなさいごめんなさいはしたなくてごめんなさい。も、もうだめです、もう……こ、こんな、はしたない格好して……せんぱいに見せ付けて……
+dataset/asu/fem_asu_12130.wav|2|あっ……せ、せんぱいっ……んっ……あぁ……それ、だめ、ですっ……あっ、あぁ……おく、当たってます……んああっ……ぐりぐり、しないでっ……
+dataset/asu/fem_asu_12331.wav|2|カメラで撮られてるのにっ、せんぱいとするの気持ちよすぎて、イクッ、イッちゃうっ……あっ、ああっ、んっ、らめっ、あっ、あああっ
+dataset/kan/fem_kan_00061.wav|1|『まだ』あんたが原因とは言ってない、って言ったのよ！
+dataset/hir/man_hir_00014.wav|6|今日の運勢１位は～～～……蟹座！　調子にのらず慎ましく生きようねっ☆
+dataset/kan/fem_kan_00180.wav|1|じゃあ、改めて。私は常磐かの、２年生。よろしくね
+dataset/kan/fem_kan_00231.wav|1|こうなってくると、和泉は動物ではなく、植物として認識されている可能性が……？
+dataset/shi/fem_shi_00033.wav|3|なぜ私が、今後関わるかもわからない人間に配慮をしなければならない……？
+dataset/shi/fem_shi_00140.wav|3|私は思いたったら行方不明になる。秘境駅を訪ねてるかもしれないし、無人島へ行っているかもしれない。急な連絡には応えかねるから、何事も早めの行動を心がけろ
+dataset/rir/fem_rir_00005.wav|7|しおさんは以前から、私たちとの話しあいの結果を無視して、独断で動かれる方でしたので、一度厳しく注意しようと考えていたところだったんですよ？
+dataset/ame/fem_ame_00085.wav|4|そだっけ！？　あ、あれ～？　昨日もここでかいちょーと話したよね？　あはは……
+dataset/kan/fem_kan_00502.wav|1|あんた、まだふざけた気持ちでいるでしょ？　私が怒ってるの、軽く考えてない？　私をなんだと思ってんの？
+dataset/asu/fem_asu_00315.wav|2|また、少しだけ大切に扱ってもらえると嬉しい……です
+dataset/kan/fem_kan_00676.wav|1|あんたが傷を負ってもいいなら、お願いしたいかも……ありがと
+dataset/shi/fem_shi_00505.wav|3|うぅ、私……ここでは奴隷のように扱われていて……太ももに顔を挟まれたいとか、その三浦大根みたいな足で踏まれたいとか、そんなことばかり……
+dataset/shi/fem_shi_00553.wav|3|なあに、このミッションを終えれば、ひよりんが私になついてくれる。そう思えば、どんな屈辱も苦ではない
+dataset/hiy/fem_hiy_00710.wav|0|愛しの妹には突っ張りを与えて、錦さんには慰めの言葉をかけただと……
+dataset/kan/fem_kan_00877.wav|1|いっ、いや、そんなにかわいいを連呼されても、私は絶対……！　身内でかわいい言いあってお互いを慰めあう痛いレイヤーになりたくない……！
+dataset/mir/fem_mir_00280.wav|5|現地の人からは『地元を楽しんでもらえれば、それが何よりも生活の張りになります』って言われてるし
+dataset/kan/fem_kan_00973.wav|1|男の子の部屋初めて……参考資料によく覚えていきたい
+dataset/asu/fem_asu_00556.wav|2|非常識なのはわかってるんです。わかってるん���すけど……ミリ先生が、おかしなことではないと言っていたので……！
+dataset/hiy/fem_hiy_00903.wav|0|あ、卒業したら、お父さんたちの遺産が使えるようになるのかー。そうしたら、ほんとに、いつ一人暮らしを始めてもおかしくなくなっちゃうね
+dataset/ame/fem_ame_00340.wav|4|ごめんもう上げちった。事後でごめん、はいこれ
+dataset/hiy/fem_hiy_01029.wav|0|お米ないから麺類で！　それなら簡単、すぐできるよ！
+dataset/hiy/fem_hiy_10126.wav|0|あー、やっぱりそうなんだー。私が浅はかだった、考えが甘かったよー
+dataset/hiy/fem_hiy_10297.wav|0|でも今、お兄に冷たい言葉をかけられたり、よそよそしい態度をされたら、心が折れてしまいそうなので……
+dataset/hiy/fem_hiy_10388.wav|0|お風呂も一緒に入ろうねー！　うちのお風呂そんなに広くないから、身体が密着してしまうねー！
+dataset/hiy/fem_hiy_10514.wav|0|夕ごはんのことしか聞かれなかったから、もしかしてお兄は知らないのかなって……そんなわけないかあ
+dataset/hiy/fem_hiy_10688.wav|0|お兄といちゃついてる感覚を味わいたいので、あーんして
+dataset/asu/fem_asu_10042.wav|2|私たちも舞台袖に待機してるからね、ひよりん。なにかあったらすぐ駆けつけるよ
+dataset/shi/fem_shi_10113.wav|3|不本意ではあるが、ひよりんが動けないのであれば、私もしばらく手を貸そう。和泉君は１時間経ったら、きちんと仕事へ戻るように
+dataset/hiy/fem_hiy_11112.wav|0|オフィシャルになってくると、しお先輩への執筆料が発生したり、スタジオ代がかかったりするので、事務所の判断次第だけど……
+dataset/hiy/fem_hiy_11282.wav|0|うひゃああああああああ恥ずかしいいぃー！
+dataset/hiy/fem_hiy_00407.wav|0|あの、お兄が起きないかなってドキドキしながら、手でさすったこともあった……
+dataset/hiy/fem_hiy_11590.wav|0|おにい好きっ！　おにい好きぃ……大好き……やんっ！　やあっ……や、あ、あんっ！　ああッ、はっ、ふあっ、ひうっ……やあんっ！
+dataset/hiy/fem_hiy_11797.wav|0|ふおおおお兄に痛いを思いをさせてしまうとは……ごめんねごめんね、私が慣れていないせいで……ううぅ
+dataset/kan/fem_kan_10149.wav|1|おっ……終わらせないとね！？
+dataset/kan/fem_kan_10269.wav|1|あ！　徹夜したあとの夜だもんね、寝落ちして当然だった……えと、ごめん……
+dataset/kan/fem_kan_10466.wav|1|まあ、今は……大抵のことは欲目で見逃してあげるから、ちゃんとドキドキさせてね
+dataset/kan/fem_kan_10568.wav|1|その配置に不満があるのは、ＴＬの呟きとか、実際に会ったときの空気で、なんとなく伝わってきて……けど私は、他に頼れる人がいなくて
+dataset/kan/fem_kan_10768.wav|1|あっ、ぅんッ……！　んッ、ああっ、あッ、やっ……やっ、気持ちいッ……やんっ！　あっ、やだ、足の指太くてごつごつして……気持ちいいよおっ！
+dataset/kan/fem_kan_10934.wav|1|お、おはよぅ
+dataset/asu/fem_asu_10090.wav|2|あ、あの。しお先輩に助けていただくことはできませんか？　和泉先輩の負担を少しでも軽くできれば……
+dataset/shi/fem_shi_10184.wav|3|すまない錦さん、その場所を私と代わってくれないか……和泉君にかぶせて、よいしょ、と
+dataset/kan/fem_kan_11277.wav|1|あっでも！　しゃわ、シャワー……シャワーだけ浴びたい
+dataset/kan/fem_kan_11431.wav|1|みんなが絵で描いてるのとも違わない……？　小さくて、芋虫みたいな棒きれ描く人いるけど、もっと大きくて複雑……
+dataset/kan/fem_kan_11631.wav|1|いじわる……もっと、今みたいなのして欲しい……
+dataset/kan/fem_kan_00390.wav|1|うわめっちゃカチカチに大きくなってる……なんか充血してるみたいで痛そう。撫でてあげるね？
+dataset/kan/fem_kan_11933.wav|1|やっ、それって舌っ……き、汚いよっ……やあっ！　あ、ああッ、あ、やっ……力、抜けちゃう……
+dataset/kan/fem_kan_12134.wav|1|やっ、ひゃふっ！　ひあっ、あんっ！　やっ、キス、したいのにいっ！　気持ち、よすぎて、出し入れやめられないっ……！
+dataset/shi/fem_shi_10389.wav|3|ごもっともだ。私は少しも悪いとは思ってないからな
+dataset/shi/fem_shi_10550.wav|3|実に充実した４日間だった……あと、もう何日か続けたかった
+dataset/kan/fem_kan_12342.wav|1|まあいいんじゃない？　書きたいものを書くのが一番でしょ
+dataset/mir/fem_mir_10368.wav|5|つかごめん、マジごめん。うちずっと嘘ついててさ。生徒会長決めるときにトモすけが引いたくじ引きのアレ、あれ全部うちの仕込みなんだ
+dataset/shi/fem_shi_10812.wav|3|自分で言うのもなんだが、私の感想はキツかっただろう？
+dataset/shi/fem_shi_10945.wav|3|ああでもせっかくだから、今のうちにすることを済ませておこうか。出せるかな？
+dataset/shi/fem_shi_11074.wav|3|マイダーリンはカフェオレがいいらしい。それならダーリンの分の湯呑みはいらないな。皆にだけ配ろう
+dataset/shi/fem_shi_11251.wav|3|さ、動いてみてくれ。この温かい感触が、いずれ快感に変わるのなら楽しみだ
+dataset/shi/fem_shi_11452.wav|3|そうだな……すっかり君に甘えてしまっていた。心地よくてつい……
+dataset/shi/fem_shi_11652.wav|3|くっ、私に、前立腺はない、のに……んッ！　はあっ、君に、ここまで開発されるとは思わなかった……
+dataset/shi/fem_shi_11850.wav|3|まさか……

filelists/hamidashi_val_filelist.txt.cleaned

@@ -0,0 +1,73 @@
+dataset/hiy/fem_hiy_12064.wav|0|a↑re↓are o↓aniga i↓ʦuni na↓kudaraʃinai ka↑ode ni↓koniko ʃI↑te i↑ru↓zo? na↓ni mi↑te↓ru no↑kana? wa↑taʃino ga↑zoo?
+dataset/asu/fem_asu_10300.wav|2|w a, ha↓i……! a↑ri↓gatoo go↑zaima↓sU.
+dataset/hiy/fem_hiy_12116.wav|0|da↑ijo↓obu. yu↑me↓janaiyo. ko↑no hi↑yoriga ho↑ʃoo su↑ruyo.
+dataset/asu/fem_asu_10635.wav|2|se↑Npai……?
+dataset/asu/fem_asu_10789.wav|2|fU↑to↓okoono hI↑to↓taʧini, de↑te ku↓ru no↑mo wa↑ru↓kunaiyoQte ʦU↑taeru ta↑me↓no mo↑no↓dato o↑mo↓Qte i↑ru↓node……
+dataset/asu/fem_asu_10912.wav|2|bu↑itiiyuubiiiiaaruwa hI↑to↓ʦuno ka↑nooseena N↓desU. se↑Qkee↓ʃIkiwa, so↑reo o↑oeN ʃI↑ta↓idesU.
+dataset/kan/fem_kan_12215.wav|1|k i, ki↑no se↓iyo……
+dataset/hir/man_hir_10106.wav|6|da↑iyaku…… a↓mano yo↑ore↓maga de↓ru wa↓keniwa i↑kimase↓Nka?
+dataset/asu/fem_asu_11205.wav|2|i↓e, na↑rutosaNga jo↑seeno ka↑ta↓o hI↑to↓ri te↓hai ʃI↑te ku↑reta↓node, so↑no hI↑toto i↑Qʃoni i↑Qte ki↑ma↓sU.
+dataset/asu/fem_asu_11328.wav|2|de↓mo bu↓taini ta↓Qte wa↑ka↓Qta N↓desU. o↑oeN…… se↑eeNQte, to↑Qtemo u↑reʃi↓i N↓desU. gu↓NguN ʧI↑kara↓ga wa↑ite ki↑ma↓sU.
+dataset/asu/fem_asu_11523.wav|2|e↑Qto, so↑rewa, so↑no…… se↑Npaiga, ki↑moʧiyo↓ku, su↑ru↓kara……
+dataset/asu/fem_asu_11723.wav|2|ʧu↑ruru, N↓Q, ʧ u…… ʧu↓pu, ju↑ru, re↑ru…… ju↑ru…… ju↓pu…… j u ru↓ru, N…… j u pU↑pu, N↓u…… ju↓pu, j u pU↑pupu……
+dataset/asu/fem_asu_11926.wav|2|go↑meNnasaigomeNnasa↓ihaʃItanakUte go↑meNnasa↓i. m o, mo↓o da↑me↓desU, mo↓o…… k o, ko↑Nna, ha↑ʃItana↓i ka↑Qkoo ʃI↑te…… se↑Npaini mi↑seʦUkete……
+dataset/asu/fem_asu_12130.wav|2|a↓Q…… s e, se↑NpaiQ…… N↓Q…… a↓a…… so↑re, da↑me, d e su↓Q…… a↓Q, a↓a…… o↑ku, a↑taQtema↓sU…… N↑a↓aQ…… gu↓riguri, ʃi↑na↓ide Q……
+dataset/asu/fem_asu_12331.wav|2|ka↓merade to↑rarete↓ru no↑niQ, se↑Npaito su↑ru n o ki↑moʧiyo↓sugite, i ku↓Q, i↑Q ʧa↓uQ…… a↓Q, a↓aQ, N↓Q, ra↑me↓Q, a↓Q, a↑a↓aQ.
+dataset/kan/fem_kan_00061.wav|1|{ ma↓da} a↓Ntaga ge↑NiNtowa i↑Qtenai, Q↑te i↑Qta↓noyo!
+dataset/hir/man_hir_00014.wav|6|kyo↓ono u↓Nsee i↑ʧii↓wa~~~…… ka↑niza! ʧo↑oʃini no↑ra↓zu ʦU↑ʦumaʃi↓kuikiyooneQ.
+dataset/kan/fem_kan_00180.wav|1|ja↓a, a↑rata↓mete. wa↑taʃiwa to↑kiwakano, ni↑ne↓Nsee. yo↑roʃIkune.
+dataset/kan/fem_kan_00231.wav|1|ko↓o na↓Qte ku↓ruto, i↓zumiwa do↑obuʦudewanaku, ʃo↑ku↓buʦUto ʃI↑te ni↑NʃIki sa↑rete i↑ru ka↑nooseega……?
+dataset/shi/fem_shi_00033.wav|3|na↓ze wa↑taʃiga, ko↑Ngo ka↑kawa↓rukamo wa↑kara↓nai ni↑NgeNni ha↓iryoo ʃi↑nakere↓ba na↑ra↓nai……?
+dataset/shi/fem_shi_00140.wav|3|wa↑taʃiwa o↑moita↓Qtara yu↑kue fu↑meeni na↓ru. hI↑kyoo↓ekio ta↑zunete↓rukamo ʃi↑renaiʃi, mu↑jiNtooe i↑Qte i↑rukamo ʃi↑renai. kyu↑una re↑Nrakuniwa ko↑taekane↓rukara, na↑nigotomo ha↑yameno ko↑odooo ko↑korogake↓ro.
+dataset/rir/fem_rir_00005.wav|7|ʃ i o↑saNwa i↓zeNkara, wa↑taʃi↓taʧItono ha↑naʃiaino ke↑Qkao mu↓ʃi ʃI↑te, do↑kudaNde u↑go↓kareru ho↓odeʃItanode, i↑ʧido ki↑biʃi↓kU ʧu↓ui ʃi↑yooto ka↑Nga↓ete i↑ta to↑korodaQta N↓desUyo?
+dataset/ame/fem_ame_00085.wav|4|so↑daQ↓ke!? a, a↑re~? ki↑no↓omo ko↑ko de↑ka↓i ʧo↓oto ha↑na↓ʃItayone? a↑haha……
+dataset/kan/fem_kan_00502.wav|1|a↓Nta, ma↓da fu↑zake↓ta ki↑moʧide i↑ru↓deʃo? wa↑taʃiga o↑koQte↓runo, ka↑rukUkaNgaete↓nai? wa↑taʃio na↓Ndato o↑mo↓Qte N↓no?
+dataset/asu/fem_asu_00315.wav|2|ma↑ta, sU↑ko↓ʃidake ta↑iseʦuni a↑ʦUkaQte mo↑raeruto u↑reʃi↓i…… de↓sU.
+dataset/kan/fem_kan_00676.wav|1|a↓Ntaga ki↑zuo o↓Qtemo i↓inara, o↑negai ʃI↑ta↓ikamo…… a↑ri↓gato.
+dataset/shi/fem_shi_00505.wav|3|u↓u, wa↑taʃi…… ko↑kodewa do↑reeno yo↓oni a↑ʦUkawarete i↑te…… fU↑tomomoni ka↑oo ha↑samareta↓itoka, so↑no mi↑ura da↑ikoNmi↓taina a↑ʃi↓de fu↑mareta↓itoka, so↑Nna ko↑to↓bakari……
+dataset/shi/fem_shi_00553.wav|3|na↓ani, ko↑no mi↓QʃoNo o↑ere↓ba, hi↑yoriNga wa↑taʃini na↑ʦu↓ite ku↑reru. so↑o o↑mo↓eba, do↓Nna kU↑ʦujokumo ku↓dewa na↓i.
+dataset/hiy/fem_hiy_00710.wav|0|a↓iʃino i↑mooto↓niwa ʦu↑Q↓hario a↑taete, ni↓ʃIkisaNniwa na↑gusameno ko↑toba↓o ka↑ke↓tadato……
+dataset/kan/fem_kan_00877.wav|1|i↑Q, i↑ya, so↑Nnani ka↑wai↓io re↓Nko sa↑retemo, wa↑taʃiwa ze↑Qtai……! mi↑uʧide ka↑wai↓iiiaQte o↑tagaio na↑gusamea↓u i↑ta↓i re↓iyaani na↑ritaku↓nai……!
+dataset/mir/fem_mir_00280.wav|5|ge↓Nʧino hI↑tokarawa{ ji↑motoo ta↑noʃi↓Nde mo↑raere↓ba, so↑rega na↓niyorimo se↑ekaʦuno ha↑rini na↑rima↓sU} Q↑te i↑warete↓ruʃi.
+dataset/kan/fem_kan_00973.wav|1|o↑toko↓nokono he↑ya ha↑ji↓mete…… sa↑Nkooʃi↓ryooni yo↓ku o↑boe↓te i↑kIta↓i.
+dataset/asu/fem_asu_00556.wav|2|hi↑jo↓oʃIkina no↑wa wa↑kaQte↓ru N↓desU. wa↑kaQte↓ru N↓desUkedo…… mi↑riseNse↓ega, o↑ka↓ʃina ko↑to↓dewa na↓ito i↑Qte i↑ta↓node……!
+dataset/hiy/fem_hiy_00903.wav|0|a, so↑ʦugyoo ʃI↑ta↓ra, o↑toosaN↓taʧino i↑saNga ʦU↑kae↓ru yo↓oni na↓ru no↑kaa. so↑oʃIta↓ra, ho↑Ntoni, i↓ʦUhitoriguraʃio ha↑jimetemo o↑kaʃi↓kunaku na↑Qʧau↓ne.
+dataset/ame/fem_ame_00340.wav|4|go↑meN mo↓o a↑geʧiQta. ji↓gode go↑meN, ha↑i↓kore.
+dataset/hiy/fem_hiy_01029.wav|0|o↓kome na↓ikara me↓Nruide! so↑rena↓ra ka↑NtaN, su↓gu de↑ki↓ruyo!
+dataset/hiy/fem_hiy_10126.wav|0|a↓a, ya↑Qpa↓ri so↑o na↓N da↓a. wa↑taʃiga a↑sahakadaQta, ka↑Nga↓ega a↑ma↓kaQtayoo.
+dataset/hiy/fem_hiy_10297.wav|0|de↓mo i↓ma, o↓anini ʦu↑metai ko↑toba↓o ka↑kerareta↓ri, yo↑soyosoʃi↓i ta↓idoo sa↑reta↓ra, ko↑ko↓roga o↑re↓te ʃi↑mai↓soonanode……
+dataset/hiy/fem_hiy_10388.wav|0|o↓furomo i↑Qʃoni ha↓iroonee! u↑ʧino o↓furo so↑Nnani hi↑ro↓kunaikara, ʃi↓Ntaiga mi↑Qʧaku ʃI↑te ʃi↑maune↓e!
+dataset/hiy/fem_hiy_10514.wav|0|yu↑ugo↓haNno ko↑to↓ʃIka kI↑karenakaQta↓kara, mo↓ʃIkaʃIte o↓aniwa ʃi↑ranai no↑kanaQte…… so↑Nna wa↑ke↓naika a.
+dataset/hiy/fem_hiy_10688.wav|0|o↓anito i↑ʧaʦuite↓ru ka↑Nkakuo a↑jiwaita↓inode, a↓aN ʃI↑te.
+dataset/asu/fem_asu_10042.wav|2|wa↑taʃi↓taʧimo bu↑tai↓sodeni ta↓iki ʃI↑te↓rukarane, hi↑yoriN. na↓nika a↓Qtara su↓gu ka↑keʦUkeruyo.
+dataset/shi/fem_shi_10113.wav|3|fu↑ho↓Nidewa a↓ruga, hi↑yoriNga u↑goke↓nai no↑deare↓ba, wa↑taʃimo ʃi↑ba↓rakU te↓o ka↑soo. i↓zumikuNwa i↑ʧiji↓kaN ta↓Qtara, kI↑ʧi↓Nto ʃi↑gotoe mo↑do↓ru yo↓oni.
+dataset/hiy/fem_hiy_11112.wav|0|o↑fIʃaruni na↓Qte ku↓ruto, ʃ i o↑se↓Npaieno ʃi↑QpIʦu↓ryooga ha↑Qsee ʃI↑ta↓ri, sU↑tajio↓daiga ka↑kaQta↓ri su↑ru↓node, ji↑mu↓ʃono ha↓NdaN ʃi↑daida↓kedo……
+dataset/hiy/fem_hiy_11282.wav|0|u↑hyaaaaaaa↓aahazukaʃiiiii!
+dataset/hiy/fem_hiy_00407.wav|0|a↑no, o↓aniga o↑ki↓naikanaQte do↓kidokI ʃi↑na↓gara, te↓de sa↑suQta ko↑to↓mo a↓Qta……
+dataset/hiy/fem_hiy_11590.wav|0|o↓nii sU↑ki↓Q! o↑niisUkii…… da↓isUki…… y a N↓Q! y a a↓Q…… y a, a, a↓NQ! a↓aQ, ha↓Q, fu↑a↓Q, hi↑uQ…… ya↓aNQ!
+dataset/hiy/fem_hiy_11797.wav|0|fu↑o↓oo o↓anini i↑ta↓io o↑mo↓io sa↑sete ʃi↑mautowa…… go↑meNne go↑meNne, wa↑taʃiga na↑re↓te i↑nai se↓ide…… u↑uu.
+dataset/kan/fem_kan_10149.wav|1|o↑Q…… o↑warasenaitone!?
+dataset/kan/fem_kan_10269.wav|1|a! te↑ʦuya ʃI↑ta a↓tono yo↓ruda mo↓Nne, ne↑oʧi ʃI↑te to↑ozeNdaQta…… e↑to, go↑meN……
+dataset/kan/fem_kan_10466.wav|1|ma↓a, i↓mawa…… ta↑iteeno ko↑to↓wa yo↑kume↓de mi↑nogaʃIte a↑geru↓kara, ʧa↑Nto do↓kidokI sa↑setene.
+dataset/kan/fem_kan_10568.wav|1|so↑no ha↑iʧini fu↑maNga a↓ru no↑wa, ti↓i e↓ru n o ʦu↑bu↓yakItoka, ji↑Qsaini a↓Qta to↓kino ku↓ukide, na↑Ntona↓kU ʦu↑tawaQte ki↓te…… ke↓do wa↑taʃiwa, ta↓ni ta↑yore↓ru ji↑Ngai na↓kUte.
+dataset/kan/fem_kan_10768.wav|1|a↓Q, u↓NQ……! N↓Q, a↓aQ, a↓Q, ya↓Q…… ya↓Q, ki↑moʧi↓iQ…… y a N↓Q! a↓Q, ya↓da, a↑ʃi↓no yu↑bi fU↑to↓kUte go↓ʦugoʦu ʃI↑te…… ki↑moʧii↓iyo o↑Q!
+dataset/kan/fem_kan_10934.wav|1|o, o↑hayou.
+dataset/asu/fem_asu_10090.wav|2|a, a↑no. ʃ i o↑se↓Npaini ta↑sUke↓te i↑tadakU ko↑to↓wa de↑kimase↓Nka? i↓zumi se↑Npaino fU↑taNo sU↑ko↓ʃidemo ka↑ruku↓dekireba……
+dataset/shi/fem_shi_10184.wav|3|su↑ma↓nai ni↓ʃIkisaN, so↑no ba↑ʃoo wa↑taʃIto ka↑waQte ku↑renai↓ka…… i↓zumikuNni ka↑buse↓te, yo↓iʃo, t o.
+dataset/kan/fem_kan_11277.wav|1|a↓Q de↓mo! ʃa↑wa, ʃa↓waa…… ʃa↑waadake a↑bita↓i.
+dataset/kan/fem_kan_11431.wav|1|mi↑Nna↓ga e↓de e↑gaite↓ru no↑tomo ʧi↑gawanai……? ʧi↑isa↓kUte, i↑mo↓muʃimitaina bo↑oki↓re e↑ga↓kU hi↑to i↑ru↓kedo, mo↓Qto o↓okIkute fU↑kuzaʦu……
+dataset/kan/fem_kan_11631.wav|1|i↑jiwa↓ru…… mo↓Qto, i↓ma mi↓tainano ʃI↑tehoʃii……
+dataset/kan/fem_kan_00390.wav|1|u↑wa me↓Qʧa ka↑ʧIkaʧini o↑okIkunaQte↓ru…… na↑Nkaju↓ukeʦu ʃI↑te↓ru mi↓taide i↑ta↓soo. na↑de↓te a↑gerune?
+dataset/kan/fem_kan_11933.wav|1|ya↓Q, so↑reQte ʃI↑ta↓Q…… k i, kI↑tana↓iyo Q…… y a a↓Q! a, a↓aQ, a, ya↓Q…… ʧI↑kara, nu↑keʧau……
+dataset/kan/fem_kan_12134.wav|1|ya↓Q, hy a fu↓Q! hi↓aQ, a↓NQ! ya↓Q, ki↓su, ʃI↑ta↓inoni i↑Q! ki↑moʧi, yo↓sugite, da↑ʃi↓ire ya↑merare↓naiQ……!
+dataset/shi/fem_shi_10389.wav|3|go↑moQto↓moda. wa↑taʃiwa sU↑koʃimo wa↑ru↓itowa o↑moQte↓naikarana.
+dataset/shi/fem_shi_10550.wav|3|ji↑ʦu↓ni ju↑ujiʦu ʃI↑ta yo↑Qka↓kaNdaQta…… a↓to, mo↓o na↑N↓niʧIka ʦu↑zuketa↓kaQta.
+dataset/kan/fem_kan_12342.wav|1|ma↓a i↓i N↓janai? ka↑kIta↓i mo↑no↓o ka↓ku no↑ga i↑ʧi↓baNdeʃo.
+dataset/mir/fem_mir_10368.wav|5|ʦU↑kago↓meN, ma↑jigo↓meN. u↑ʧi zu↑Qto u↓so ʦu↓itetesa. se↑etoka↓iʧoo ki↑meru to↓kini to↓mo sU↑kega hi↑ita ku↑ji↓bikino a↑re, a↑re ze↓Nbu u↑ʧino ʃI↑komina N↓da.
+dataset/shi/fem_shi_10812.wav|3|ji↑buNde i↑u no↑mo na↓Ndaga, wa↑taʃino ka↑Nsoowa kI↑ʦu↓kaQtadaroo?
+dataset/shi/fem_shi_10945.wav|3|a↓a de↓mo se↑Qkaku da↓kara, i↓mano u↑ʧini su↑ru ko↑to↓o su↑mase↓te o↓kooka. da↑se↓rukana?
+dataset/shi/fem_shi_11074.wav|3|ma↑ida↓ariNwa ka↑feorega i↓iraʃii. so↑rena↓ra da↓ariNno bu↓Nno yu↑nomi↓wa i↑ranai↓na. mi↑nanidake ku↑ba↓roo.
+dataset/shi/fem_shi_11251.wav|3|s a, u↑go↓ite mi↓te ku↑re. ko↑no a↑tataka↓i ka↑Nʃokuga, i↑zureka↓ikaNni ka↑waru no↑nara ta↑noʃi↓mida.
+dataset/shi/fem_shi_11452.wav|3|so↑oda↓na…… su↑Qka↓ri ki↑mini a↑maete ʃi↑ma↓Qte i↑ta. ko↑koʧiyo↓kUte ʦu↓i……
+dataset/shi/fem_shi_11652.wav|3|ku↓Q, wa↑taʃini, ze↑NriʦUseNwa na↓i, no↓ni…… N↓Q! w a a↓Q, ki↑mini, ko↑koma↓de ka↑ihaʦU sa↑rerutowa o↑mowa↓nakaQta……
+dataset/shi/fem_shi_11850.wav|3|ma↓saka……

filelists/train_filelist.txt

@@ -0,0 +1,291 @@
+louise/VOICE_ID_04514.wav|サ～イ～ト～！
+louise/VOICE_ID_04515.wav|わたしがちょおっと目を離したスキに、メイドに妙な格好させてニヤついてるとは、いい度胸ねぇサイト……。
+louise/VOICE_ID_04516.wav|逃がさないわよサイト！
+louise/VOICE_ID_04517.wav|許さないっ！
+louise/VOICE_ID_04518.wav|まったくあんたって使い魔は、どうしてメイドやらツェルプストーの娘やらにやたらと色目を使うかしらねぇ……。
+louise/VOICE_ID_04519.wav|犬だって3日飼えばご主人様への忠誠を身につけるというのに、あんたって全くもって犬以下ね。
+louise/VOICE_ID_04520.wav|どうやら、今回は徹底的にしつけをしなくちゃいけないようね。
+louise/VOICE_ID_04521.wav|おだまり！もとはといえば、全部サイトが悪いのよ。
+louise/VOICE_ID_04522.wav|こうなったら徹底的にお仕置きしてやるわ！自分の立場をきっちり思い出させてあげる。
+louise/VOICE_ID_04523.wav|そ、そうね。お、お風呂でご主人様の背中を流すってどうかしら？
+louise/VOICE_ID_04524.wav|ど、どう？屈辱でしょ？お仕置きでしょ？
+louise/VOICE_ID_04525.wav|な、な、なんですって！お、お仕置きっていいなさい。怖がりなさい！
+louise/VOICE_ID_04526.wav|わ、分かったのなら、一緒にお風呂に行くわよ。わたしは準備するから、そこで待ってなさい。
+louise/VOICE_ID_04527.wav|ほら、何ブツブツ言っているの？早くお仕置きするわよ。
+louise/VOICE_ID_04528.wav|この水着は、ハルナからセンベツでもらったものよ。
+louise/VOICE_ID_04529.wav|サイトが言うことを聞かなかったら、これを使ってくださいって、言ってた。
+louise/VOICE_ID_04530.wav|な、何言ってるの？水着を着ないでサイトと一緒にお風呂に入れるはずないでしょ？
+louise/VOICE_ID_04531.wav|ほら。
+louise/VOICE_ID_04532.wav|あぁもうじれったいわね！
+louise/VOICE_ID_04533.wav|それでわたしの背中を流しなさいって言ってるの。あ、でもその前に……。
+louise/VOICE_ID_04534.wav|もう一枚タオルがあるでしょ？それでぎっちり目隠しをするのを忘れないでね。
+louise/VOICE_ID_04535.wav|背中を洗うとき、少し背中を見せるけど、もし、ゆる～く目隠しなんかしたら……。コ・ロ・スからね。
+louise/VOICE_ID_04536.wav|ん、どれどれ……。ふぅん、ちゃんと言いつけは守ったようね。
+louise/VOICE_ID_04537.wav|よっと。ほら、準備はできたわ。さっさとわたしの背中を洗いなさい！
+louise/VOICE_ID_04538.wav|そうねぇ、もうちょっと強くしてもいいわよ。
+louise/VOICE_ID_04539.wav|あー、そこそこ……。うん、そこがいいわ。もっと、重点的にね。
+louise/VOICE_ID_04540.wav|ふんふんふーん。ねぇ、サイト。今は、どんな心境かしら？
+louise/VOICE_ID_04541.wav|ふふん、負け惜しみは言わなくていいわよ。お風呂場の中でわたしと2人っきり、しかも背中を流してるのに……、
+louise/VOICE_ID_04542.wav|なんにも見えないってのはつらいでしょ？もどかしいでしょ？正直、拷問よねぇ。
+louise/VOICE_ID_04543.wav|まぁ、それこそが今回のお仕置きの真の目的なんだけどね。
+louise/VOICE_ID_04544.wav|これに懲りたらもう、わたしをないがしろになんかしないことね。
+louise/VOICE_ID_04545.wav|ちょっとサイト、わたしの話聞いてるの？手もさっきから止まってるわよ！
+louise/VOICE_ID_04546.wav|まったく、ちょっとほめたらすぐいい気になって……。これだからサイトは。
+louise/VOICE_ID_04547.wav|あー、気持ちいい。サイト、あなた体洗うの上手だわ。
+louise/VOICE_ID_04548.wav|んー、それにしても……。はぁ……。おっきくならないのよねぇ。
+louise/VOICE_ID_04549.wav|な、なんだっていいでしょ。手が止まってる！
+louise/VOICE_ID_04550.wav|ちょっとサイト、あんたさっきから様子がヘンよ？まさか湯あたりでもしたの？
+louise/VOICE_ID_04551.wav|もうちょっといろんなところが大きくなればいいのに……。サイト、さっきからまた手が止まってる？
+louise/VOICE_ID_04552.wav|きゃ、きゃあぁっ！サイト？どうしたのよあんた、倒れちゃって。
+louise/VOICE_ID_04553.wav|ちょ……サイト！大丈夫？
+louise/VOICE_ID_04554.wav|目を開けなさいって！サイト、サイト、サイトぉぉぉぉぉ！！！！！
+louise/VOICE_ID_04639.wav|なに？なにか用？
+louise/VOICE_ID_04640.wav|なに？どうしたのよ。そんなところで固まって。
+louise/VOICE_ID_04641.wav|ちょっと、サイト！
+louise/VOICE_ID_04642.wav|なに？じゃないわよ。なにか言うこ���はないの？
+louise/VOICE_ID_04643.wav|もう！
+louise/VOICE_ID_04644.wav|使い魔だったら、ご主人様がいつもと違った格好してるって気づいたら、何か意見くらい言いなさいよね！
+louise/VOICE_ID_04645.wav|え？な、なによ急に……。
+louise/VOICE_ID_04646.wav|しょ、正直に……？そ、そうなの。あ……ありがとう。
+louise/VOICE_ID_04647.wav|う……、いいのっ！あんたは黙ってなさい！
+louise/VOICE_ID_04648.wav|なによそれ！わたしがこういう服を着ちゃ、似合わないってこと！？
+louise/VOICE_ID_04649.wav|え、ええと……？
+louise/VOICE_ID_04650.wav|つまり、普段のわたしっぽくないって言いたいわけ！？
+louise/VOICE_ID_04651.wav|せっかく、新しい服を見せてあげたのに、そういうことしか言えないの、あんたは！
+louise/VOICE_ID_04652.wav|……は？
+louise/VOICE_ID_04653.wav|な、なにかあるでしょ。可愛いとか、きれいとか、似合ってるとか、そういうのが。
+louise/VOICE_ID_04654.wav|うー……もういいわよ！この鈍感！
+louise/VOICE_ID_04655.wav|知らない！いいから出て行きなさい！
+louise/VOICE_ID_04892.wav|やれやれじゃないわよ。あんたが休むのは、仕事を終わらせてからよ！
+louise/VOICE_ID_04893.wav|誰が奥さんよ！
+louise/VOICE_ID_04894.wav|服は洗ったの？部屋の掃除はした？買ってきた物はちゃんとしまった？
+louise/VOICE_ID_04895.wav|だったら、さっさと片づけなさいよ！終わるまで休憩なしだからねっ！
+louise/VOICE_ID_04896.wav|何か文句あるの！？
+louise/VOICE_ID_04897.wav|……。
+louise/VOICE_ID_04898.wav|あの……あの、ね？ちょっと、聞こうかなって思ったんだけどね？
+louise/VOICE_ID_04899.wav|サイト……元の世界に帰りたい？
+louise/VOICE_ID_04900.wav|……そう。そう、だよね……やっぱり。
+louise/VOICE_ID_04901.wav|なんでもない！
+louise/VOICE_ID_04902.wav|けど？
+louise/VOICE_ID_04903.wav|ふーん。
+louise/VOICE_ID_04904.wav|……噓つき。
+louise/VOICE_ID_04905.wav|元いた世界に帰りたくないだなんて、そんなの嘘に決まってるじゃない。
+louise/VOICE_ID_04906.wav|すぐばれるような噓をつくもんじゃないわ。それって、すごく失礼だって分かるでしょ。
+louise/VOICE_ID_04907.wav|もういいわよ。
+louise/VOICE_ID_04908.wav|さあて、と。あんたの用事はまだ終わってないんだから、さっさと始めなさい！
+louise/VOICE_ID_04909.wav|うるさいうるさい！早く仕事するの！
+louise/VOICE_ID_04910.wav|なあに？何を聞きたいの？
+louise/VOICE_ID_04911.wav|どこかの誰かさんが挙動不審だったから、気になって部屋に戻ってみたのよ。戻ってきて正解だったみたいね！
+louise/VOICE_ID_04912.wav|ご主人様が授業に出てる隙に、いったい何をしようとしていたのかしら？
+louise/VOICE_ID_04913.wav|へー、そうなの？
+louise/VOICE_ID_04914.wav|使い魔にしては、ずいぶん立派な心がけね。
+louise/VOICE_ID_04915.wav|でも、わたしに黙ってやることでもないわよねぇ……。
+louise/VOICE_ID_04916.wav|ふーん。まぁ、今日はそういうことにしてあげる。
+louise/VOICE_ID_04917.wav|ふーん。で、何を聞きたいわけ？わたしの目の前で聞いてみなさいよ。
+louise/VOICE_ID_04918.wav|ほら、わたしに聞かれたくないこと、聞こうと思ってたんでしょ？
+louise/VOICE_ID_04919.wav|問答無用ー！
+louise/VOICE_ID_04920.wav|は？
+louise/VOICE_ID_04921.wav|なによ、それ。もう少しまともな理由を考えなさいよ！
+louise/VOICE_ID_04922.wav|はあ……こんなのを相手に本気で怒るのも、バカみたいだわ。
+louise/VOICE_ID_04923.wav|ねえ？
+louise/VOICE_ID_04924.wav|まあいいわ。今から授業に戻れないし、みんなが来るまで、ここで待ちましょう。
+louise/VOICE_ID_04925.wav|何か文句ある？
+louise/VOICE_ID_05539.wav|……。
+louise/VOICE_ID_05540.wav|なによ。
+louise/VOICE_ID_05541.wav|……言ってごらんなさい。
+louise/VOICE_ID_05542.wav|そんなこともあったかしらね。
+louise/VOICE_ID_05543.wav|そりゃあ、あんたは楽しかったでしょうね。そこらじゅうの女の子にデレデレしちゃって……。
+louise/VOICE_ID_05544.wav|もう、みっともないったらありゃしない。
+louise/VOICE_ID_05545.wav|！！！！！そ、そ、そ、そんなわけないでしょうが。
+louise/VOICE_ID_05546.wav|バカ使い魔がそこらじゅうに迷惑をかけたら、ご主人様が迷惑するからよ！別にヤキモチやいてるわけじゃないわよ！
+louise/VOICE_ID_05547.wav|そうよ！
+louise/VOICE_ID_05548.wav|い、いいじゃない！わたしがオシャレしたら、なにが悪い！？
+louise/VOICE_ID_05549.wav|あ、そ、そう……？
+louise/VOICE_ID_05550.wav|うー、なによなによ、さっきから文句ばっかり！あんた、わたしの使い魔でしょ？文句言わないで働きなさいよ！
+louise/VOICE_ID_05551.wav|第一、せっかくわたしが店で服を試着しても、まともに見ないし感想も言ってくれないし。
+louise/VOICE_ID_05552.wav|なんでもないわよ！最近は、ちゃんとご飯食べさせてるんだし、ご主人様の言うことくらい素直に聞きなさい！
+louise/VOICE_ID_05553.wav|なにか言った！？
+louise/VOICE_ID_05554.wav|敷いてない！
+louise/VOICE_ID_05555.wav|それ以上減らず口叩いたら、溶かして屑鉄にして学院の裏庭に埋めるからね！
+louise/VOICE_ID_05556.wav|ちょっと、サイト？どうしたの？
+louise/VOICE_ID_05557.wav|え？どこ？
+louise/VOICE_ID_05558.wav|あ、ちょっと、サイト！もうっ、何なのよー。
+louise/VOICE_ID_05559.wav|もう、サイト！ご主人様を置いて何してるのよ！って、本当に人が倒れてる……。
+louise/VOICE_ID_05560.wav|この娘、見慣れない格好をしてるけど、一体、どこの国の人かしら。
+louise/VOICE_ID_05561.wav|ちょっとサイト。なに、この娘をじーっと見つめてるのよ。
+louise/VOICE_ID_05562.wav|サイト、なにやってんのよ！女の子の服を脱がそうとするなんて！
+louise/VOICE_ID_05563.wav|全然そうには見えないわよ！わたしが診るからあんたはあっちに行って。
+louise/VOICE_ID_05564.wav|意識を失ってるだけにようね。頭に損傷は無いみたい……。サイト、急いで学院に向かって！
+louise/VOICE_ID_05565.wav|「へっ」じゃないわよ。わたし達は馬なんだから、倒れてる人を連れていけるわけないでしょ？
+louise/VOICE_ID_05566.wav|応援を呼んでって意味よ。まさか、最後まで話さないと分からないってことはないでしょ？
+louise/VOICE_ID_05567.wav|なっ！？なに、あんた達！
+louise/VOICE_ID_05568.wav|サイト……。
+louise/VOICE_ID_05569.wav|ちょっとサイト！どこの誰だか分からない連中にこの娘を預けちゃうつもり！？
+louise/VOICE_ID_05570.wav|ふーん？
+louise/VOICE_ID_05571.wav|何、考えこんでるのよ！この場合は断固拒否、でしょ！
+louise/VOICE_ID_05572.wav|分かってるわよ！こんなことに巻き込んで……。サイト、後で覚えてなさいよ！
+louise/VOICE_ID_05573.wav|逃がすもんですか！追うわよ、サイト！
+louise/VOICE_ID_05574.wav|ちょっと、あいつらを放っておく気？
+louise/VOICE_ID_05575.wav|……まあ、そうかもしれないけど。でも、なーんか気分が収まらないわね。
+louise/VOICE_ID_05576.wav|しつこい！
+louise/VOICE_ID_05577.wav|今度やったら、本当に割るからね。それより、サイト。これからどうするの？
+louise/VOICE_ID_05578.wav|あんな連中が狙う理由なり事情があるってことは、ただの娘じゃないんでしょ？わたし達の手に負えるとは限らないわ。
+louise/VOICE_ID_05579.wav|一度街まで引き返して、この娘を保護してもらうって手もあるでしょ？
+louise/VOICE_ID_05580.wav|……なんで、そんなにこの娘にこだわるのよ。
+louise/VOICE_ID_05581.wav|あやしい！サイト。この娘に何をしたのか、白状しなさい！
+louise/VOICE_ID_05582.wav|初めて会った割にはずいぶんとご執心だし、大きい胸はあんたの好みだし、そのうえ、あのメイドと同じ黒い髪だし！
+louise/VOICE_ID_05583.wav|さーあ、白状なさい！この娘にいったい何をしたの！？
+louise/VOICE_ID_05584.wav|……本当に？じゃあなんでそんなに熱心なのよ。
+louise/VOICE_ID_05585.wav|あ……。
+louise/VOICE_ID_05586.wav|……。
+louise/VOICE_ID_05587.wav|もうっ、分かったわよ！元々そのつもりだったわけだし、いったん学院に連れて行きましょ。
+louise/VOICE_ID_05588.wav|さっきの人達がまた襲ってくるかもしれないし、この娘はわたしの馬に乗せるわ。こっちにつけてる荷物そっちに移すわね。
+louise/VOICE_ID_05589.wav|……その言い方が、信用できないっていうのよ。真面目に話す気がないのね。そうなのね？
+louise/VOICE_ID_05590.wav|問答無用ー！
+louise/VOICE_ID_05591.wav|はあ、はあ、はあ……。
+louise/VOICE_ID_05592.wav|まあ、いいわ。元々そのつもりだったわけだし、いったん学院に連れて行きましょ。
+louise/VOICE_ID_05593.wav|あんたのためじゃないわよ。あんたがこの娘にちょっかい出したら、今度は手加減なしだからね。
+louise/VOICE_ID_05594.wav|さっきの人達がまた襲ってくるかもしれないし、この娘はわたしの馬に乗せるわ。こっちにつけてる荷物そっちに移す���ね。
+louise/VOICE_ID_05595.wav|でも、事情が事情だけど、平民を勝手に学院に入れるのって先生達に知られるのもまずいわよね、多分。
+louise/VOICE_ID_05596.wav|手当ての必要はあるんだけど、どうしたらいいのかしら。
+louise/VOICE_ID_05597.wav|まったく、なんでわたしがこんなこと……。
+louise/VOICE_ID_05598.wav|それより、シエスタ。いきなり部屋に病人を担ぎ込んじゃってすまなかったわね。
+louise/VOICE_ID_05599.wav|……ほおお。
+louise/VOICE_ID_05600.wav|え。
+louise/VOICE_ID_05601.wav|ええと、そ、そう！今日、街に出たときに会ったんだけど、どうやらわたしの故郷から来た平民らしいの。
+louise/VOICE_ID_05602.wav|途中で財布を盗られたとかで、行くところがないって言うから、しばらくの間面倒を見てあげようと思って。
+louise/VOICE_ID_05603.wav|モンモランシー、わたしからもお願い！秘密にしてくれないかしら？
+louise/VOICE_ID_05604.wav|あとは、彼女が目を覚ましてくれればいいんだけど……。どうしたものかしらね？
+louise/VOICE_ID_05605.wav|本当に、サイトのせいで余計な面倒をかけちゃうけど、よろしくね。
+louise/VOICE_ID_05606.wav|ほら、サイト！せっかくシエスタがああ言ってくれてるんだから、部屋に戻るわよ！
+louise/VOICE_ID_05607.wav|ちょっとサイト。なに、部屋の隅に行ってんのよ。
+louise/VOICE_ID_05608.wav|す、隅っこでいいなら、一緒にベッドを使わせてあげるわよ！
+louise/VOICE_ID_05609.wav|いいの！わたしがそう言ってるんだから、さっさとベッドに入りなさいよ。
+louise/VOICE_ID_05610.wav|何をブツブツ言ってるのよ。ほら、さっさと寝るわよ！
+louise/VOICE_ID_05611.wav|ちょっと……あんまりもぞもぞ動かないでよ。
+louise/VOICE_ID_05612.wav|別に……少しくらい、いいけど。
+louise/VOICE_ID_05613.wav|な、なによ……急に変なこと言い出して。
+louise/VOICE_ID_05614.wav|そうよ。当たり前のこと言わないでよ。
+louise/VOICE_ID_05615.wav|もう……変なの。
+louise/VOICE_ID_05616.wav|……ねえ、サイト？起きてる？
+louise/VOICE_ID_05617.wav|寝ちゃった……の？
+louise/VOICE_ID_05618.wav|……もう！いいわよ、わたしも寝ちゃうから！
+louise/VOICE_ID_05619.wav|ううーん……。
+louise/VOICE_ID_05620.wav|ん……ふわああ……。あれ、サイト……？
+louise/VOICE_ID_05621.wav|……って、ええっ！？なんでシエスタがわたしの部屋にいるのよっ！
+louise/VOICE_ID_05622.wav|ん、んー……。
+louise/VOICE_ID_05623.wav|ううーん……。
+louise/VOICE_ID_05624.wav|ん……ふわああ……。あれ、サイト……？
+louise/VOICE_ID_05625.wav|……って、ええっ！？あんたなんでわたしに抱きついてんのよーっ！
+louise/VOICE_ID_05626.wav|もう、そのことはいいでしょ、別に！それより、なんでシエスタがわたしの部屋にいるのよっ！
+louise/VOICE_ID_05627.wav|んん……。
+louise/VOICE_ID_05628.wav|ふわああ……。もう、朝からうるさいわよサイト。おかげで目が覚めちゃったじゃ……。
+louise/VOICE_ID_05629.wav|……って、ええっ！？なんでシエスタがわたしの部屋にいるのよっ！
+louise/VOICE_ID_05630.wav|え、あ、ちょっとサイト！？わたしが着替えるまで待ちなさいよ！！
+louise/VOICE_ID_05631.wav|なんで、朝からこんなに大勢集まってるのよ！
+louise/VOICE_ID_05632.wav|そんなの決まってるじゃない。ねえ、モンモランシー？
+louise/VOICE_ID_05633.wav|別に楽しいことなんてないわよ。
+louise/VOICE_ID_05634.wav|ってことは、答えはひとつよね。ねえ、モンモランシー？
+louise/VOICE_ID_05635.wav|別に楽しいことなんてないわよ。
+louise/VOICE_ID_05636.wav|あんたと一緒に来たわたしに、そんなこと分かるわけないでしょ！
+louise/VOICE_ID_05637.wav|ああ、それはそうね。でも、原因は分かってるわ。
+louise/VOICE_ID_05638.wav|ねえ、シエスタ。あなたがわたし達を呼びに来る前は、こんなに賑やかじゃなかったんでしょ？
+louise/VOICE_ID_05639.wav|ってことは、答えはひとつよね。ねえ、モンモランシー？
+louise/VOICE_ID_05640.wav|まあ、いいわ。何を言ったところで別に帰るとは思えないし。
+louise/VOICE_ID_05641.wav|それより、この娘が誰なのかが問題なの。
+louise/VOICE_ID_05642.wav|ちなみに、ベッドに寝かせたのはシエスタで、容態を診たのはモンモランシーよ。
+louise/VOICE_ID_05643.wav|で、今度はこちらが聞きたいんだけど。あなた、どこから来たの？名前は？
+louise/VOICE_ID_05644.wav|やっぱりサイトが知っている娘だった���じゃない！
+louise/VOICE_ID_05645.wav|まあいいわ。で、あなたはどうやってここに来たの？
+louise/VOICE_ID_05646.wav|けど？
+louise/VOICE_ID_05647.wav|ふうん？ずいぶんと剣呑な話ね。
+louise/VOICE_ID_05648.wav|知ってる人もいないし、行く当てもないし、そのまま放浪していたら、倒れちゃったってところかしら。
+louise/VOICE_ID_05650.wav|な、なに、なんでいきなりわたしの使い魔に抱きつくのよ！離れなさい！
+louise/VOICE_ID_05651.wav|うっ！
+louise/VOICE_ID_05652.wav|外野は黙ってて！
+louise/VOICE_ID_05653.wav|ええと、すごーく特別な例なんだけど、サイトはわたしが使い魔を召喚する魔法で呼び出して、契約を交わしたの。
+louise/VOICE_ID_05654.wav|だから、サイトは人間だけどわたしの使い魔なの。ほら、左手に使い魔の印があるでしょ。
+louise/VOICE_ID_05655.wav|なんですってえ！？
+louise/VOICE_ID_05656.wav|い、いけなくはないけど、表現方法に問題が……。
+louise/VOICE_ID_05657.wav|な……！
+louise/VOICE_ID_05658.wav|な、な、な……！？
+louise/VOICE_ID_05659.wav|……サイト。お願いだから今すぐその女達から離れなさい。
+louise/VOICE_ID_05660.wav|ふ、ふふふ、ふふふ……。そう、サイトはあくまで開き直るつもりなのね？
+louise/VOICE_ID_05661.wav|なんですってえ！？
+louise/VOICE_ID_05662.wav|ふ、ふふふ、ふふふ……。そう、サイトはあくまで開き直るつもりなのね？
+louise/VOICE_ID_05663.wav|分かった！！！だったら、力ずくで離してあげるわっ！！
+louise/VOICE_ID_05664.wav|問答無用ー！！
+louise/VOICE_ID_05665.wav|……。
+louise/VOICE_ID_05666.wav|は、はい！ええと、その……。
+louise/VOICE_ID_05667.wav|魔法の実験中、つい魔法が暴走してしまい、爆発してしまいました。
+louise/VOICE_ID_05668.wav|サイト！？なにか、間違いでも！？
+louise/VOICE_ID_05669.wav|す、すみません。
+louise/VOICE_ID_05670.wav|は、はい！ありがとうございます！
+louise/VOICE_ID_05671.wav|ちょっとサイト！何を言いだすのよ！
+louise/VOICE_ID_05672.wav|それとこれとは話が違うでしょ！だいたい、あんたの部屋じゃなくてわたしの部屋でしょうが！
+louise/VOICE_ID_05673.wav|何を冷静に答えてるのよ！部屋の主はわたし！あんたは使い魔なの！
+louise/VOICE_ID_05674.wav|使い魔が勝手に部屋に仲間を連れこまれないで！
+louise/VOICE_ID_05675.wav|し、しかし……！
+louise/VOICE_ID_05676.wav|……分かりました。
+louise/VOICE_ID_05677.wav|ちょっとキュルケ！何を言いだすのよ！
+louise/VOICE_ID_05678.wav|それとこれとは話が違うでしょ！
+louise/VOICE_ID_05679.wav|説得力はあるけど、だったら余計にお断りよ！
+louise/VOICE_ID_05680.wav|し、しかし……！
+louise/VOICE_ID_05681.wav|ちょ、ちょっと、オールド・オスマン！それはいくらなんでも、そのっ！
+louise/VOICE_ID_05682.wav|え、ええと……。
+louise/VOICE_ID_05683.wav|……分かりました。
+louise/VOICE_ID_05684.wav|……。
+louise/VOICE_ID_05685.wav|……ちょっと、ね。不幸な事故があったの。
+louise/VOICE_ID_05686.wav|くっ……！
+louise/VOICE_ID_05687.wav|ところでモンモランシー？改めて尋ねたいことがあるんだけど。
+louise/VOICE_ID_05688.wav|……なんで目線をそらすのかしら。
+louise/VOICE_ID_05689.wav|ちゃんと黙っててって言ったでしょ！なんでこんなにばれまくってるのよ！
+louise/VOICE_ID_05690.wav|つい、じゃないわよ、もう。秘密って言ったのにこれじゃ、先が思いやられるわ。
+louise/VOICE_ID_05691.wav|モンモランシー。わたし、同情するわ。
+louise/VOICE_ID_05692.wav|そうね。じゃあ、わたしは授業があるから、終わったら全員で集合ってことでいい？
+louise/VOICE_ID_05693.wav|さて、みんな集まったわね。
+louise/VOICE_ID_05694.wav|仕方ないわよ。モンモランシーがばらしちゃったんだもの。この際、手を貸してもらうわ。
+louise/VOICE_ID_05695.wav|あら、大丈夫よ。この娘には、すぐに出てってもらうから。
+louise/VOICE_ID_05696.wav|なによ。
+louise/VOICE_ID_05697.wav|なにがひどいのよ。わたしは当たり前のことを言ってるだけよ。
+louise/VOICE_ID_05698.wav|なんとかって、なによ。
+louise/VOICE_ID_05699.wav|平民1人をわたし達だけでかくまえって言うの？そんなの無理だわ。
+louise/VOICE_ID_05700.wav|元々は、この娘が道ばたで倒れてたから介抱するために学院まで連れてきたわけでしょ。
+louise/VOICE_ID_05701.wav|なら部外者が学院にいるのはおかしいの。
+louise/VOICE_ID_05702.wav|わたし、なにか間違ったこと言って��？
+louise/VOICE_ID_05703.wav|それじゃあ、わたし達でこの先どうやって彼女の面倒を見てくの？
+louise/VOICE_ID_05704.wav|犬や猫じゃないのよ。そんなこと、無理に決まってるじゃない。
+louise/VOICE_ID_05705.wav|は？
+louise/VOICE_ID_05706.wav|な……っ！そ、そ、そんなんじゃないわよ！
+louise/VOICE_ID_05707.wav|そんなはずない！なんで、このサイトに嫉妬なんて、しなくちゃならないのよ！
+louise/VOICE_ID_05708.wav|う……。
+louise/VOICE_ID_05709.wav|分かったわよ！当分の間かくまえばいいんでしょ！
+louise/VOICE_ID_05710.wav|襲ってきた連中が、実は彼女を召喚した魔法使いかその部下ってこと？
+louise/VOICE_ID_05711.wav|あ、そうなの？
+louise/VOICE_ID_05712.wav|サイト？あんた、今エッチなこと考えたでしょ。
+louise/VOICE_ID_05713.wav|断言はできないけど、そういうことになるわね。
+louise/VOICE_ID_05714.wav|……。
+louise/VOICE_ID_05715.wav|もう起きてるわよ。
+louise/VOICE_ID_05716.wav|まったく、ご主人様を放ったらかしにしてぐーぐー寝てるなんて、いったい何様のつもりかしら？
+louise/VOICE_ID_05717.wav|う、うるさいわね。本当はこのくらいに起きてるのよ。
+louise/VOICE_ID_05718.wav|そっ、そんなわけないでしょ。ちゃんと寝たわよ。
+louise/VOICE_ID_05719.wav|だから、ちゃんと寝てるって言ってるじゃない！それにいつも寝てるのはあんたでしょ？
+louise/VOICE_ID_05720.wav|ふん、やる気が足りないのよ。
+louise/VOICE_ID_05721.wav|う……ち、違うわよ。部屋の人口密度が高いから、ちょっと寝苦しかっただけよ。
+louise/VOICE_ID_05722.wav|それってどういう意味よ！
+louise/VOICE_ID_05723.wav|本当なの？まぁ、それならいいんだけどね。
+louise/VOICE_ID_05724.wav|へ？なんで？
+louise/VOICE_ID_05725.wav|なんで、わたしが使い魔の行動を逐次監視しなくちゃ駄目なのよ！第一、サイトは何かをするつもりだったの？
+louise/VOICE_ID_05726.wav|そう？ならいいけど。
+louise/VOICE_ID_05727.wav|もし、ちゃっかいを出していたのなら、それ相応の罰を受けてもらうけどね。
+louise/VOICE_ID_05728.wav|ふふん、わたしよりよっぽどお寝坊さんがいたみたいね。
+louise/VOICE_ID_05729.wav|ちょ、ちょっと何謝ってるのよ！しっかりしなさいよ。
+louise/VOICE_ID_05730.wav|それより、そろそろ授業に行かないといけない時間だわ。
+louise/VOICE_ID_05731.wav|いいから！さっさと授業に行くわよ
+louise/VOICE_ID_05732.wav|あ……！！すぐ着替えするから、部屋から出てって！
+louise/VOICE_ID_05733.wav|早く部屋から出てって！着替えはシエスタにしてもらうから。
+louise/VOICE_ID_05734.wav|ぐずぐずしてないで、さっさと出てって！
+louise/VOICE_ID_05735.wav|もう、サイトが早く部屋から出てくれなかったから遅れたんでしょ？
+louise/VOICE_ID_05736.wav|え……えぇ！？あ、いえ、その、なんでもありません。ミスタ・コルベール。
+louise/VOICE_ID_05737.wav|はい。
+louise/VOICE_ID_05738.wav|ま、わたし達には関係ないけどね。

filelists/train_filelist.txt.cleaned

@@ -0,0 +1,291 @@
+louise/VOICE_ID_04514.wav|s a~ i~ t o~!
+louise/VOICE_ID_04515.wav|wa↑taʃiga ʧ o o↓Qto me↓o ha↑na↓ʃIta sU↑kini, me↓idoni myo↓ona ka↑Qkoo sa↑sete ni↓ya ʦu↓iterutowa, i↓i do↓kyoonee sa↑ito…….
+louise/VOICE_ID_04516.wav|ni↑gasa↓naiwayo sa↑ito!
+louise/VOICE_ID_04517.wav|yu↑rusa↓naiQ!
+louise/VOICE_ID_04518.wav|ma↑Qtaku a↓NtaQte ʦU↑kai ma↑wa, do↓oʃIte me↓idoyara ʦe↑rupUsu↓toono mu↑sume↓yarani ya↑tarato i↑ro↓meo ʦU↑kau↓kaʃiranee…….
+louise/VOICE_ID_04519.wav|i↑nu↓daQte mi↑Qka ka↓eba go↑ʃu↓jiNsamaeno ʧu↑useeo mi↑ni ʦU↑ke↓ruto i↑u↓noni, a↓NtaQte ma↑Qtaku mo↓Qte i↑nui↓kane.
+louise/VOICE_ID_04520.wav|do↓oyara, ko↓Nkaiwa te↑Qteetekini ʃI↑ʦukeo ʃi↑nakUʧa i↑kenai yo↓one.
+louise/VOICE_ID_04521.wav|o↓da ma↑ri! mo↑to↓wato i↓eba, ze↓Nbu sa↑itoga wa↑ru↓inoyo.
+louise/VOICE_ID_04522.wav|ko↓o na↓Qtara te↑Qteetekini o↑ʃioki ʃI↑te ya↑ru↓wa! ji↑buNno ta↓ʧibao ki↑Qʧi↓ri o↑moida↓sasete a↑geru.
+louise/VOICE_ID_04523.wav|s o, so↑one. o, o↓furode go↑ʃu↓jiNsamano se↑nakao na↑ga↓suQte do↓okaʃira?
+louise/VOICE_ID_04524.wav|d o, do↓o? kU↑ʦujokudeʃo? o↑ʃiokideʃo?
+louise/VOICE_ID_04525.wav|n a, n a, na↓NdesUQte! o, o↑ʃiokiQte i↓inasai. ko↑waga↓rinasai!
+louise/VOICE_ID_04526.wav|w a, wa↑ka↓Qta no↑nara, i↑Qʃoni o↓furoni i↑ku↓wayo. wa↑taʃiwa ju↓Nbi su↑ru↓kara, so↑kode ma↓Qte na↑sa↓i.
+louise/VOICE_ID_04527.wav|ho↓ra, na↑Nbu↓ʦubuʦu i↑Qte i↑ru↓no? ha↓yaku o↑ʃiokI su↑ru↓wayo.
+louise/VOICE_ID_04528.wav|ko↑no mi↑zugiwa, ha↓runakara se↑Nbeʦude mo↑raQta mo↑no↓yo.
+louise/VOICE_ID_04529.wav|sa↑itoga i↑u ko↑to↓o kI↑kanakaQta↓ra, ko↑reo ʦU↑kaQte ku↑dasa↓iQte, i↑Qteta.
+louise/VOICE_ID_04530.wav|n a, na↓ni i↑Qte↓runo? mi↑zugio ki↓naide sa↑itoto i↑Qʃoni o↓furoni i↑reru ha↑zu na↓ideʃo?
+louise/VOICE_ID_04531.wav|ho↓ra.
+louise/VOICE_ID_04532.wav|a↓a mo↓o ji↑reQta↓iwane!
+louise/VOICE_ID_04533.wav|so↑rede wa↑taʃino se↑nakao na↑gaʃi↓nasaiQte i↑Qte↓runo. a, de↓mo so↑no ma↓eni…….
+louise/VOICE_ID_04534.wav|mo↓o i↑ʧi↓mai ta↓oruga a↓rudeʃo? so↑rede gi↑Qʧi↓ri me↑ka↓kUʃio su↑ru no↑o wa↑surena↓idene.
+louise/VOICE_ID_04535.wav|se↑nakao a↑rau to↓ki, sU↑ko↓ʃI se↑nakao mi↑se↓rukedo, mo↓ʃi, yu↑ru~ ku↑meka↓kUʃinaNka ʃI↑ta↓ra……. k or osu↓karane.
+louise/VOICE_ID_04536.wav|N, do↑redore……. fu↓uN, ʧa↑Nto i↑iʦUke↓wa ma↑mo↓Qta yo↓one.
+louise/VOICE_ID_04537.wav|yo↑Qto. ho↓ra, ju↓Nbiwa de↑ki↓tawa. sa↓Qsato wa↑taʃino se↑nakao a↑rainasai!
+louise/VOICE_ID_04538.wav|so↑onee, mo↓o ʧo↓Qto ʦu↑yo↓kuʃItemo i↓iwayo.
+louise/VOICE_ID_04539.wav|a↓a, so↑kosoko……. u↓N, so↑koga i↓iwa. mo↓Qto, ju↑uteNtekinine.
+louise/VOICE_ID_04540.wav|fu↑NfuNfu↓uN. ne↓e, sa↑ito. i↓mawa, do↓Nna ʃi↑Nkyookaʃira?
+louise/VOICE_ID_04541.wav|fu↓fuN, ma↑keoʃimiwa i↑wanakUteii↓wayo. o↑fu↓robano na↓kade wa↑taʃIto fU↑tari Q↑kiri, ʃI↑ka↓mo se↑nakao na↑gaʃIte↓runoni……,
+louise/VOICE_ID_04542.wav|na↑Nnimo mi↑e↓naiQte no↑wa ʦu↑ra↓ideʃo? mo↑dokaʃi↓ideʃo? ʃo↑oji↓ki, go↑omoNyonee.
+louise/VOICE_ID_04543.wav|ma↓a, so↑reko↓soga ko↓Nkaino o↑ʃiokino ʃi↓Nno mo↑kUtekina N↓dakedone.
+louise/VOICE_ID_04544.wav|ko↑reni ko↑ri↓tara mo↓o, wa↑taʃio na↑iga↓ʃironi na↑Nkaʃi↓nai ko↑to↓ne.
+louise/VOICE_ID_04545.wav|ʧo↓Qto sa↑ito, wa↑taʃino ha↑naʃi ki↑ite↓runo? te↓mo sa↓QkIkara to↑maQte↓ruwayo!
+louise/VOICE_ID_04546.wav|ma↑Qtaku, ʧo↓Qto ho↑me↓tara su↓gu i↓i ki↑ni na↓Qte……. ko↑reda↓kara sa↑itowa.
+louise/VOICE_ID_04547.wav|a↓a, ki↑moʧii↓i. sa↑ito, a↑natata↓i a↑rau n o jo↑ozu↓dawa.
+louise/VOICE_ID_04548.wav|N↑N, so↑reniʃIte↓mo……. ha↓a……. o↑QkIkunara↓nainoyonee.
+louise/VOICE_ID_04549.wav|n a, na↓NdaQte i↓ideʃo. te↓ga to↑maQte↓ru!
+louise/VOICE_ID_04550.wav|ʧo↓Qto sa↑ito, a↓Nta sa↓QkIkara yo↑osuga he↓Nyo? ma↓saka yu↑a↓taridemo ʃI↑ta↓no?
+louise/VOICE_ID_04551.wav|mo↓o ʧo↓Qto i↑roNna to↑koroga o↓okIkunareba i↓i no↑ni……. sa↑ito, sa↓QkIkara ma↑ta te↓ga to↑maQte↓ru?
+louise/VOICE_ID_04552.wav|ky a, kya↓aaQ! sa↑ito? do↓o ʃI↑ta↓noyo a↓Nta, ta↑ore↓ʧaQte.
+louise/VOICE_ID_04553.wav|ʧ o…… sa↑ito! da↑ijo↓obu?
+louise/VOICE_ID_04554.wav|me↓o a↑kenasaiQte! sa↑ito, sa↑ito, sa↑itoooooo!!!!!
+louise/VOICE_ID_04639.wav|na↓ni? na↑nika↓yoo?
+louise/VOICE_ID_04640.wav|na↓ni? do↓o ʃI↑ta↓noyo. so↑Nna to↑korode ka↑tamaQte.
+louise/VOICE_ID_04641.wav|ʧo↓Qto, sa↑ito!
+louise/VOICE_ID_04642.wav|na↓ni? j a na↓iwayo. na↓nika i↑u ko↑to↓wanaino?
+louise/VOICE_ID_04643.wav|mo↓o!
+louise/VOICE_ID_04644.wav|ʦU↑kai↓madaQtara, go↑ʃu↓jiNsamaga i↓ʦumoto ʧi↑gaQta ka↑Qkoo ʃI↑te↓ruQte ki↑zu↓itara, na↓nika i↑keNku↓rai i↑inasaiyone!
+louise/VOICE_ID_04645.wav|e? n a, na↓niyo kyu↑uni…….
+louise/VOICE_ID_04646.wav|ʃ o, ʃo↑oji↓kini……? s o, so↑ona↓no. a…… a↑ri↓gatoo.
+louise/VOICE_ID_04647.wav|u……, i↓i no↓Q! a↓Ntawa da↑ma↓Qte na↑sa↓i!
+louise/VOICE_ID_04648.wav|na↓niyo so↑re! wa↑taʃiga ko↑oyuu fU↑ku↓o ki↓ʧa, ni↑awa↓naiQte ko↑to!?
+louise/VOICE_ID_04649.wav|e, e↓eto……?
+louise/VOICE_ID_04650.wav|ʦu↓mari, fu↓daNno wa↑taʃiQpokuna↓iQte i↑ita↓i wa↓ke!?
+louise/VOICE_ID_04651.wav|se↑Qkaku, a↑taraʃi↓i fU↑ku↓o mi↑se↓te a↑ge↓tanoni, so↑oyuu ko↑to↓ʃIka i↑enai↓no, a↓Ntawa!
+louise/VOICE_ID_04652.wav|…… w a?
+louise/VOICE_ID_04653.wav|n a, na↓nika a↓rudeʃo. ka↑wai↓itoka, ki↓reitoka, ni↑aQte↓rutoka, so↑oyuu no↑ga.
+louise/VOICE_ID_04654.wav|u↓u…… mo↓o i↓iwayo! ko↑no do↑NkaN!
+louise/VOICE_ID_04655.wav|ʃi↑ranai! i↓ikara de↑te i↑kinasai!
+louise/VOICE_ID_04892.wav|ya↓reyarejanaiwayo. a↓Ntaga ya↑su↓mu no↑wa, ʃi↑gotoo o↑warasetekarayo!
+louise/VOICE_ID_04893.wav|da↓rega o↓kUsaNyo!
+louise/VOICE_ID_04894.wav|fU↑ku↓wa a↑raQta↓no? he↑ya↓no so↑ojiwa ʃI↑ta? ka↑Qte ki↓ta mo↑no↓wa ʧa↑Nto ʃi↑ma↓Qta?
+louise/VOICE_ID_04895.wav|da↑Qta↓ra, sa↓Qsato ka↑tazuke↓nasaiyo! o↑waruma↓de kyu↑ukee na↓ʃidakaraneQ!
+louise/VOICE_ID_04896.wav|na↓nika mo↓Nku a↓runo!?
+louise/VOICE_ID_04897.wav|…….
+louise/VOICE_ID_04898.wav|a↑no…… a↑no, n e? ʧo↓Qto, kI↑koo↓kanaQte o↑mo↓Qta N↓dakedone?
+louise/VOICE_ID_04899.wav|sa↑ito…… mo↑to↓no se↓kaini ka↑erita↓i?
+louise/VOICE_ID_04900.wav|…… so↑o. so↑o, da↑yone…… ya↑Qpa↓ri.
+louise/VOICE_ID_04901.wav|na↓Ndemo na↓i!
+louise/VOICE_ID_04902.wav|ke↓do?
+louise/VOICE_ID_04903.wav|fu↑uN.
+louise/VOICE_ID_04904.wav|…… u↑so↓ʦUki.
+louise/VOICE_ID_04905.wav|mo↑to↓ita se↓kaini ka↑eritaku↓naidanaNte, so↑Nna no↓usoni ki↑maQte↓rujanai.
+louise/VOICE_ID_04906.wav|su↓gu ba↑re↓ru yo↓ona u↓soo ʦu↓ku mo↓Nja na↓iwa. so↑reQte, su↑go↓kU ʃi↑ʦu↓reedaQte wa↑ka↓rudeʃo.
+louise/VOICE_ID_04907.wav|mo↓o i↓iwayo.
+louise/VOICE_ID_04908.wav|sa↓ate, t o. a↓Ntano yo↑ojiwa ma↓da o↑waQtenai N↓dakara, sa↓Qsato ha↑jimenasai!
+louise/VOICE_ID_04909.wav|u↑rusa↓iurusai! ha↓yakU ʃi↑goto su↑ru↓no!
+louise/VOICE_ID_04910.wav|na↓ani? na↓nio kI↑kita↓ino?
+louise/VOICE_ID_04911.wav|do↓kokano da↓rekasaNga kyo↑doo fU↑ʃiNdaQta↓kara, ki↑ni na↓Qte he↑ya↓ni mo↑do↓Qte mi↓tanoyo. mo↑do↓Qte ki↓te se↑ekaidaQta mi↓taine!
+louise/VOICE_ID_04912.wav|go↑ʃu↓jiNsamaga ju↓gyooni de↑te↓ru sU↑kini, i↑Qtai na↓nio ʃi↑yooto ʃI↑te i↑ta no↑kaʃira?
+louise/VOICE_ID_04913.wav|he↑e, so↑ona↓no?
+louise/VOICE_ID_04914.wav|ʦU↑kai↓mani ʃI↑tewa, zu↓ibuN ri↑Qpana ko↑korogakene.
+louise/VOICE_ID_04915.wav|de↓mo, wa↑taʃini da↑ma↓Qte ya↑ru ko↑to↓demo na↓iwayonee…….
+louise/VOICE_ID_04916.wav|fu↑uN. ma↓a, kyo↓owa so↑oyuu ko↑to↓ni ʃI↑te a↑geru.
+louise/VOICE_ID_04917.wav|fu↑uN. d e, na↓nio kI↑kita↓i wa↓ke? wa↑taʃino me↑noma↓ede ki↑ite mi↓nasaiyo.
+louise/VOICE_ID_04918.wav|ho↓ra, wa↑taʃini kI↑karetaku↓nai ko↑to, kI↑kooto o↑mo↓Qteta N↓deʃo?
+louise/VOICE_ID_04919.wav|mo↓Ndoo mu↑yooo!
+louise/VOICE_ID_04920.wav|w a?
+louise/VOICE_ID_04921.wav|na↓niyo, so↑re. mo↑osUko↓ʃi ma↑tomona ri↑yuuo ka↑Nga↓enasaiyo!
+louise/VOICE_ID_04922.wav|ha↓a…… ko↑Nna no↑o a↑ite↓ni ho↑Nkide o↑ko↓ru no↑mo, ba↓ka mi↓taidawa.
+louise/VOICE_ID_04923.wav|ne↓e?
+louise/VOICE_ID_04924.wav|ma↓a i↓iwa. i↓makara ju↓gyooni mo↑dore↓naiʃi, mi↑Nna↓ga ku↓rumade, ko↑kode ma↑ʧimaʃo↓o.
+louise/VOICE_ID_04925.wav|na↓nika mo↓Nku a↓ru?
+louise/VOICE_ID_05539.wav|…….
+louise/VOICE_ID_05540.wav|na↓niyo.
+louise/VOICE_ID_05541.wav|…… i↑Qte go↑raN na↑sa↓i.
+louise/VOICE_ID_05542.wav|so↑Nna ko↑to↓mo a↓Qtakaʃirane.
+louise/VOICE_ID_05543.wav|so↑rya a, a↓Ntawa ta↑noʃi↓kaQtadeʃoone. so↑ko↓ra ju↓uno o↑Nna↓nokoni de↑redere ʃI↑ʧaQte…….
+louise/VOICE_ID_05544.wav|mo↓o, mi↑Qtomona↓iQtara a↑ryaʃi↓nai.
+louise/VOICE_ID_05545.wav|!!!!! s o, s o, s o, so↑Nna wa↑ke↓naideʃooga.
+louise/VOICE_ID_05546.wav|ba↑kaʦUkai↓maga so↑ko↓ra ju↓uni me↓ewakuo ka↑ke↓tara, go↑ʃu↓jiNsamaga me↓ewakU su↑ru↓karayo! be↑ʦuni ya↑kimo↓ʧi ya↑ite↓ru wa↓kejanaiwayo!
+louise/VOICE_ID_05547.wav|so↑oyo!
+louise/VOICE_ID_05548.wav|i, i↓ijanai! wa↑taʃiga o↑ʃa↓re ʃI↑ta↓ra, na↓niga wa↑ru↓i!?
+louise/VOICE_ID_05549.wav|a, s o, so↑o……?
+louise/VOICE_ID_05550.wav|u↓u, na↓niyo na↓niyo, sa↓QkIkara mo↓NkubaQkari! a↓Nta, wa↑taʃino ʦU↑kai↓madeʃo? mo↓Nku i↑wana↓ide ha↑tarakinasaiyo!
+louise/VOICE_ID_05551.wav|da↑i↓iʧi, se↑Qkaku wa↑taʃiga mi↑se↓de fU↑ku↓o ʃI↑ʧaku ʃI↑temo, ma↑tomoni mi↓naiʃI ka↑Nsoomo i↑Qte ku↑renaiʃi.
+louise/VOICE_ID_05552.wav|na↓Ndemo na↓iwayo! sa↑ikiNwa, ʧa↑Nto go↓haN ta↑besasete↓ru N↓daʃi, go↑ʃu↓jiNsamano i↑u ko↑toku↓rai su↓naoni kI↑kinasai!
+louise/VOICE_ID_05553.wav|na↓nika i↑Qta!?
+louise/VOICE_ID_05554.wav|ʃi↑itenai!
+louise/VOICE_ID_05555.wav|so↑reijoohe↓razuguʧI ta↑ta↓itara, to↑ka↓ʃIte ku↑zuteʦuni ʃI↑te ga↑kuiNno u↑raniwani u↑meru↓karane!
+louise/VOICE_ID_05556.wav|ʧo↓Qto, sa↑ito? do↓o ʃI↑ta↓no?
+louise/VOICE_ID_05557.wav|e? do↓ko?
+louise/VOICE_ID_05558.wav|a, ʧo↓Qto, sa↑ito! mo↓o Q, na↓nina no↑yoo.
+louise/VOICE_ID_05559.wav|mo↓o, sa↑ito! go↑ʃu↓jiNsamao o↑ite na↓ni ʃI↑te↓runoyo! Q↑te, ho↑Ntooni hI↑toga ta↑orete↓ru…….
+louise/VOICE_ID_05560.wav|ko↑no mu↑sume, mi↑narenai ka↑Qkooo ʃI↑te↓rukedo, i↑Qtai, do↓kono ku↑nino hI↑tokaʃira.
+louise/VOICE_ID_05561.wav|ʧo↓Qto sa↑ito. na↓ni, ko↑no mu↑sume↓o ji↑iQto mi↑ʦumete↓runoyo.
+louise/VOICE_ID_05562.wav|sa↑ito, na↓niyaQte N↓noyo! o↑Nna↓nokono fU↑ku↓o nu↑ga↓sooto su↑ru↓naNte!
+louise/VOICE_ID_05563.wav|ze↑NzeN so↑oniwa mi↑e↓naiwayo! wa↑taʃiga mi↓rukara a↓Ntawa a↑Qʧi↓ni i↑Qte.
+louise/VOICE_ID_05564.wav|i↓ʃIkio u↑ʃinaQte↓rudakeni yo↓une. a↑tama↓ni so↑Nʃoowa na↓i mi↓tai……. sa↑ito, i↑so↓ide ga↑kuiNni mu↑kaQte!
+louise/VOICE_ID_05565.wav|[ he↓Q] j a na↓iwayo. wa↑taʃi↓taʧiwa u↑ma↓na N↓dakara, ta↑orete↓ru hI↑too ʦu↑rete i↑keruwakenai↓deʃo?
+louise/VOICE_ID_05566.wav|o↑oeNo yo↑NdeQte i↓miyo. ma↓saka, sa↓igomade ha↑nasa↓naito wa↑kara↓naiQte ko↑to↓wanaideʃo?
+louise/VOICE_ID_05567.wav|na↓Q!? na↓ni, a↑Nta↓taʧi!
+louise/VOICE_ID_05568.wav|sa↑ito…….
+louise/VOICE_ID_05569.wav|ʧo↓Qto sa↑ito! do↓kono da↓redaka wa↑kara↓nai re↑Nʧuuni ko↑no mu↑sume↓o a↑zukeʧau ʦu↑mori!?
+louise/VOICE_ID_05570.wav|fu↑uN?
+louise/VOICE_ID_05571.wav|na↓ni, ka↑NgaekoNde↓runoyo! ko↑no ba↑aiwa da↓Nko kyo↓hi, de↓ʃo!
+louise/VOICE_ID_05572.wav|wa↑kaQte↓ruwayo! ko↑Nna ko↑to↓ni ma↑kIko↓Nde……. sa↑ito, a↓tode o↑boe↓te na↑sa↓iyo!
+louise/VOICE_ID_05573.wav|ni↑ga↓su mo↓NdesUka! o↑u↓wayo, sa↑ito!
+louise/VOICE_ID_05574.wav|ʧo↓Qto, a↑iʦu↓rao ha↑na↓Qte o↑kU k i?
+louise/VOICE_ID_05575.wav|…… ma↓a, so↑okamo ʃi↑renai↓kedo. de↓mo, na↓aNka ki↓buNga o↑samara↓naiwane.
+louise/VOICE_ID_05576.wav|ʃI↑ʦuko↓i!
+louise/VOICE_ID_05577.wav|ko↓Ndo ya↓Qtara, ho↑Ntooni wa↑reru↓karane. so↑reyo↓ri, sa↑ito. ko↑rekara do↓o su↑ru↓no?
+louise/VOICE_ID_05578.wav|a↑Nna re↑Nʧuuga ne↑rau ri↑yuuna↓ri ji↑jooga a↓ruQte ko↑to↓wa, ta↓dano mu↑sume↓janai N↓deʃo? wa↑taʃi↓taʧino te↓ni o↑eru↓towa ka↑gira↓naiwa.
+louise/VOICE_ID_05579.wav|i↑ʧido ma↑ʧi↓made hI↑kikaeʃIte, ko↑no mu↑sume↓o ho↓go ʃI↑te mo↑rauQte te↓mo a↓rudeʃo?
+louise/VOICE_ID_05580.wav|…… na↓Nde, so↑Nnani ko↑no mu↑sume↓ni ko↑dawa↓ru no↑yo.
+louise/VOICE_ID_05581.wav|a↑yaʃii! sa↑ito. ko↑no mu↑sume↓ni na↓nio ʃI↑ta no↑ka, ha↓kujoo ʃi↑nasai!
+louise/VOICE_ID_05582.wav|ha↑ji↓mete a↓Qta wa↑riniwa zu↓ibuNto go↑ʃu↓uʃiNdaʃi, o↑oki↓i mu↑ne↓wa a↓Ntano ko↓nomidaʃi, so↑noue, a↑no me↓idoto o↑naji ku↑ro↓i ka↑mi↓daʃi!
+louise/VOICE_ID_05583.wav|sa↑a a, ha↓kujoo na↑sa↓i! ko↑no mu↑sume↓ni i↑Qtai na↓nio ʃI↑ta↓no!?
+louise/VOICE_ID_05584.wav|…… ho↑Ntooni? ja↓a na↓Nde so↑Nnani ne↓QʃiNnanoyo.
+louise/VOICE_ID_05585.wav|a…….
+louise/VOICE_ID_05586.wav|…….
+louise/VOICE_ID_05587.wav|mo↓o Q, wa↑ka↓Qtawayo! mo↑tomoto so↑no ʦu↑moridaQta wa↓kedaʃi, i↑QtaN ga↑kuiNni ʦu↑rete i↑kimaʃo.
+louise/VOICE_ID_05588.wav|sa↓Qkino hI↑to↓taʧiga ma↑ta o↑so↓Qte ku↓rukamo ʃi↑renaiʃi, ko↑no mu↑sume↓wa wa↑taʃino u↑ma↓ni no↑seru↓wa. ko↑Qʧi↓ni ʦU↑kete↓ru ni↑moʦUso↓Qʧini u↑ʦu↓suwane.
+louise/VOICE_ID_05589.wav|…… so↑no i↑ikataga, ʃi↑Nyoo de↑ki↓naiQte i↑u↓noyo. ma↑jimeni ha↑na↓sU ki↑ga na↓inone. so↑ona↓none?
+louise/VOICE_ID_05590.wav|mo↓Ndoo mu↑yooo!
+louise/VOICE_ID_05591.wav|ha↓a, ha↓a, ha↓a…….
+louise/VOICE_ID_05592.wav|ma↓a, i↓iwa. mo↑tomoto so↑no ʦu↑moridaQta wa↓kedaʃi, i↑QtaN ga↑kuiNni ʦu↑rete i↑kimaʃo.
+louise/VOICE_ID_05593.wav|a↓Ntano ta↑me↓janaiwayo. a↓Ntaga ko↑no mu↑sume↓ni ʧo↓Qkai da↓ʃItara, ko↓Ndowa te↑ka↓geN na↓ʃi da↓karane.
+louise/VOICE_ID_05594.wav|sa↓Qkino hI↑to↓taʧiga ma↑ta o↑so↓Qte ku↓rukamo ʃi↑renaiʃi, ko↑no mu↑sume↓wa wa↑taʃino u↑ma↓ni no↑seru↓wa. ko↑Qʧi↓ni ʦU↑kete↓ru ni↑moʦUso↓Qʧini u↑ʦu↓suwane.
+louise/VOICE_ID_05595.wav|de↓mo, ji↑jooga ji↑jooda↓kedo, he↑emiNo ka↑Qteni ga↑kuiNni i↑reru↓noQte se↑Nsee↓taʧini ʃi↑rareru no↑mo ma↑zu↓iwayone, ta↓buN.
+louise/VOICE_ID_05596.wav|te↓ateno hI↑ʦuyoowa a↓ru N↓dakedo, do↓o ʃI↑ta↓ra i↓i no↑kaʃira.
+louise/VOICE_ID_05597.wav|ma↑Qtaku, na↓Nde wa↑taʃiga ko↑Nna ko↑to…….
+louise/VOICE_ID_05598.wav|so↑reyo↓ri, ʃi↑e↓sUta. i↑kinari he↑ya↓ni byo↑oniNo ka↑ʦugiko↓NjaQte su↑ma↓nakaQtawane.
+louise/VOICE_ID_05599.wav|…… h o o↓o.
+louise/VOICE_ID_05600.wav|e.
+louise/VOICE_ID_05601.wav|e↓eto, s o, so↑o! kyo↓o, ma↑ʧi↓ni de↑ta to↓kini a↓Qta N↓dakedo, do↓oyara wa↑taʃino ko↓kyookara ki↓ta he↑emiNraʃi↓ino.
+louise/VOICE_ID_05602.wav|to↑ʧuude sa↑ifuo to↑rare↓tatokade, i↑kU to↑koroga na↓iQte i↑u↓kara, ʃi↑ba↓rakuno a↑ida me↑Ndo↓oo mi↓te a↑ge↓yooto o↑mo↓Qte.
+louise/VOICE_ID_05603.wav|mo↑NmoraN↓sii, wa↑taʃIkaramo o↑negai! hi↑miʦuni ʃI↑te ku↑renai↓kaʃira?
+louise/VOICE_ID_05604.wav|a↓towa, ka↓nojoga me↓o sa↑ma↓ʃIte ku↑rere↓ba i↓i N↓dakedo……. do↓o ʃI↑ta mo↑no↓kaʃirane?
+louise/VOICE_ID_05605.wav|ho↑Ntooni, sa↑itono se↓ide yo↑keena me↑Ndo↓oo ka↑keʧau↓kedo, yo↑roʃIkune.
+louise/VOICE_ID_05606.wav|ho↓ra, sa↑ito! se↑Qkaku ʃi↑e↓sUtaga a↓aiQte ku↑rete↓ru N↓dakara, he↑ya↓ni mo↑do↓ruwayo!
+louise/VOICE_ID_05607.wav|ʧo↓Qto sa↑ito. na↓ni, he↑ya↓no su↓mini i↑Qte N↓noyo.
+louise/VOICE_ID_05608.wav|s u, su↓miQkode i↓inara, i↑Qʃoni be↓Qdoo ʦU↑kawasete a↑geru↓wayo!
+louise/VOICE_ID_05609.wav|i↓ino! wa↑taʃiga so↑o i↑Qte↓ru N↓dakara, sa↓Qsato be↓Qdoni ha↑irinasaiyo.
+louise/VOICE_ID_05610.wav|na↓nio bu↑ʦubuʦu i↑Qte↓runoyo. ho↓ra, sa↓Qsato ne↑ru↓wayo!
+louise/VOICE_ID_05611.wav|ʧo↓Qto…… a↑Nmari mo↓zomozo u↑go↓kanaideyo.
+louise/VOICE_ID_05612.wav|be↑ʦuni…… sU↑ko↓ʃIkurai, i↓ikedo.
+louise/VOICE_ID_05613.wav|n a, na↓niyo…… kyu↑uni he↓Nna ko↑to i↑ida↓ʃIte.
+louise/VOICE_ID_05614.wav|so↑oyo. a↑tarimaeno ko↑to i↑wana↓ideyo.
+louise/VOICE_ID_05615.wav|mo↓o…… he↓Nnano.
+louise/VOICE_ID_05616.wav|…… ne↓e, sa↑ito? o↑kIte↓ru?
+louise/VOICE_ID_05617.wav|ne↓ʧaQta…… n o?
+louise/VOICE_ID_05618.wav|…… mo↓o! i↓iwayo, wa↑taʃimo ne↑ʧau↓kara!
+louise/VOICE_ID_05619.wav|u↑u↓uN…….
+louise/VOICE_ID_05620.wav|N…… fu↑wa a↓a……. a↑re, sa↑ito……?
+louise/VOICE_ID_05621.wav|…… Q↑te, e↓eQ!? na↓Nde ʃi↑e↓sUtaga wa↑taʃino he↑ya↓ni i↑ru n o yo↑Q!
+louise/VOICE_ID_05622.wav|N, N↑N…….
+louise/VOICE_ID_05623.wav|u↑u↓uN…….
+louise/VOICE_ID_05624.wav|N…… fu↑wa a↓a……. a↑re, sa↑ito……?
+louise/VOICE_ID_05625.wav|…… Q↑te, e↓eQ!? a↓NtanaNde wa↑taʃini da↑kIʦu↓ite N↓noyoo Q!
+louise/VOICE_ID_05626.wav|mo↓o, so↑no ko↑to↓wa i↓ideʃo, be↑ʦuni! so↑reyo↓ri, na↓Nde ʃi↑e↓sUtaga wa↑taʃino he↑ya↓ni i↑ru n o yo↑Q!
+louise/VOICE_ID_05627.wav|N↓N…….
+louise/VOICE_ID_05628.wav|fu↑wa a↓a……. mo↓o, a↓sakara u↑rusa↓iwayo sa↑ito. o↑kagede me↓ga sa↑me↓ʧaQta j a…….
+louise/VOICE_ID_05629.wav|…… Q↑te, e↓eQ!? na↓Nde ʃi↑e↓sUtaga wa↑taʃino he↑ya↓ni i↑ru n o yo↑Q!
+louise/VOICE_ID_05630.wav|e, a, ʧo↓Qto sa↑ito!? wa↑taʃiga ki↑gae↓rumade ma↓ʧinasaiyo!!
+louise/VOICE_ID_05631.wav|na↓Nde, a↓sakara ko↑Nnani o↑oze↓e a↑ʦumaQte↓runoyo!
+louise/VOICE_ID_05632.wav|so↑Nna n o ki↑maQte↓rujanai. ne↓e, mo↑NmoraN↓sii?
+louise/VOICE_ID_05633.wav|be↑ʦuni ta↑noʃi↓i ko↑to↓naNte na↓iwayo.
+louise/VOICE_ID_05634.wav|Q↑te ko↑to↓wa, ko↑ta↓ewa hI↑to↓ʦuyone. ne↓e, mo↑NmoraN↓sii?
+louise/VOICE_ID_05635.wav|be↑ʦuni ta↑noʃi↓i ko↑to↓naNte na↓iwayo.
+louise/VOICE_ID_05636.wav|a↓Ntato i↑Qʃoni ki↓ta wa↑taʃini, so↑Nna ko↑to wa↑karuwake↓naideʃo!
+louise/VOICE_ID_05637.wav|a↓a, so↑rewa so↑one. de↓mo, ge↑NiNwa wa↑kaQte↓ruwa.
+louise/VOICE_ID_05638.wav|ne↓e, ʃi↑e↓sUta. a↑na↓taga wa↑taʃi↓taʧio yo↑bini ku↓ru ma↓ewa, ko↑Nnani ni↑gi↓yakajanakaQta N↓deʃo?
+louise/VOICE_ID_05639.wav|Q↑te ko↑to↓wa, ko↑ta↓ewa hI↑to↓ʦuyone. ne↓e, mo↑NmoraN↓sii?
+louise/VOICE_ID_05640.wav|ma↓a, i↓iwa. na↓nio i↑Qta to↑korode be↑ʦuni ka↓erutowa o↑moe↓naiʃi.
+louise/VOICE_ID_05641.wav|so↑reyo↓ri, ko↑no mu↑sume↓ga da↓rena no↑kaga mo↑Ndaina↓no.
+louise/VOICE_ID_05642.wav|ʧi↑namini, be↓Qdoni ne↑kaseta no↑wa ʃi↑e↓sUtade, yo↑odaio mi↓ta no↑wa mo↑NmoraN↓siiyo.
+louise/VOICE_ID_05643.wav|d e, ko↓Ndowa ko↑ʧiraga kI↑kita↓i N↓dakedo. a↑na↓ta, do↓kokara ki↓tano? na↑maewa?
+louise/VOICE_ID_05644.wav|ya↑Qpa↓ri sa↑itoga ʃi↑Qte i↑ru mu↑sume↓daQta N↓janai!
+louise/VOICE_ID_05645.wav|ma↓a i↓iwa. d e, a↑na↓tawa do↓o ya↓Qte ko↑koni ki↓tano?
+louise/VOICE_ID_05646.wav|ke↓do?
+louise/VOICE_ID_05647.wav|fu↑uN? zu↓ibuNto ke↑NnoNna ha↑naʃi↓ne.
+louise/VOICE_ID_05648.wav|ʃi↑Qte↓ru hI↑tomo i↑naiʃi, i↑ku a↑temo na↓iʃi, so↑no ma↑maho↓oroo ʃI↑te i↑ta↓ra, ta↑ore↓ʧaQtaQte to↑korokaʃira.
+louise/VOICE_ID_05650.wav|n a, na↓ni, na↓Nde i↑kinari wa↑taʃino ʦU↑kai↓mani da↑kIʦu↓ku no↑yo! ha↓narenasai!
+louise/VOICE_ID_05651.wav|u↓Q!
+louise/VOICE_ID_05652.wav|ga↑iyawa da↑ma↓Qtete!
+louise/VOICE_ID_05653.wav|e↓eto, su↑go↓okU to↑kubeʦuna re↓ena N↓dakedo, sa↑itowa wa↑taʃiga ʦU↑kai ma↑o ʃo↑okaN su↑ru ma↑hoode yo↑bidaʃIte, ke↑eyakuo ka↑waʃIta↓no.
+louise/VOICE_ID_05654.wav|da↓kara, sa↑itowa ni↑NgeNda↓kedo wa↑taʃino ʦU↑kai↓manano. ho↓ra, hi↑dariteni ʦU↑kai ma↑no ʃi↑ruʃiga a↓rudeʃo.
+louise/VOICE_ID_05655.wav|na↓NdesUQte e!?
+louise/VOICE_ID_05656.wav|i, i↑kenaku↓wanaikedo, hyo↑ogeNho↓ohooni mo↑Ndaiga…….
+louise/VOICE_ID_05657.wav|n a……!
+louise/VOICE_ID_05658.wav|n a, n a, n a……!?
+louise/VOICE_ID_05659.wav|…… sa↑ito. o↑negaida↓kara i↓ma su↓gu so↑no o↑Nna↓taʧIkara ha↓narenasai.
+louise/VOICE_ID_05660.wav|f u, fu↑fufu, fu↑fufu……. so↑o, sa↑itowa a↑ku↓made hi↑rakinao↓ru ʦu↑morina↓none?
+louise/VOICE_ID_05661.wav|na↓NdesUQte e!?
+louise/VOICE_ID_05662.wav|f u, fu↑fufu, fu↑fufu……. so↑o, sa↑itowa a↑ku↓made hi↑rakinao↓ru ʦu↑morina↓none?
+louise/VOICE_ID_05663.wav|wa↑ka↓Qta!!! da↑Qta↓ra, ʧI↑karazukude ha↑na↓ʃIte a↑geru↓waQ!!
+louise/VOICE_ID_05664.wav|mo↓Ndoo mu↑yooo!!
+louise/VOICE_ID_05665.wav|…….
+louise/VOICE_ID_05666.wav|w a, ha↓i! e↓eto, so↑no…….
+louise/VOICE_ID_05667.wav|ma↑hoono ji↑QkeNʧuu, ʦu↓i ma↑hooga bo↑osoo ʃI↑te ʃi↑mai, ba↑kUhaʦu ʃI↑te ʃi↑maima↓ʃIta.
+louise/VOICE_ID_05668.wav|sa↑ito!? na↓nika, ma↑ʧiga↓idemo!?
+louise/VOICE_ID_05669.wav|s u, su↑mimase↓N.
+louise/VOICE_ID_05670.wav|w a, ha↓i! a↑ri↓gatoo go↑zaima↓sU!
+louise/VOICE_ID_05671.wav|ʧo↓Qto sa↑ito! na↓nio i↑ida↓su no↑yo!
+louise/VOICE_ID_05672.wav|so↑reto ko↑retowa ha↑naʃi↓ga ʧi↑gau↓deʃo! da↑itai, a↓Ntano he↑ya↓janakUte wa↑taʃino he↑ya↓deʃooga!
+louise/VOICE_ID_05673.wav|na↓nio re↑eseeni ko↑taete↓runoyo! he↑ya↓no a↓rujiwa wa↑taʃi! a↓Ntawa ʦU↑kai ma↑na↓no!
+louise/VOICE_ID_05674.wav|ʦU↑kai↓maga ka↑Qteni he↑ya↓ni na↑kama↓o ʦu↑reko↓marenaide!
+louise/VOICE_ID_05675.wav|ʃ i, ʃI↑ka↓ʃi……!
+louise/VOICE_ID_05676.wav|…… wa↑karima↓ʃIta.
+louise/VOICE_ID_05677.wav|ʧo↓Qto kyu↓ruke! na↓nio i↑ida↓su no↑yo!
+louise/VOICE_ID_05678.wav|so↑reto ko↑retowa ha↑naʃi↓ga ʧi↑gau↓deʃo!
+louise/VOICE_ID_05679.wav|se↑Qtokuryo↓kuwa a↓rukedo, da↑Qta↓ra yo↑keeni o↑kotowariyo!
+louise/VOICE_ID_05680.wav|ʃ i, ʃI↑ka↓ʃi……!
+louise/VOICE_ID_05681.wav|ʧ o, ʧo↓Qto, o↓orudoo↓sumaN! so↑rewa i↓kura na↓Ndemo, so↑no Q!
+louise/VOICE_ID_05682.wav|e, e↓eto…….
+louise/VOICE_ID_05683.wav|…… wa↑karima↓ʃIta.
+louise/VOICE_ID_05684.wav|…….
+louise/VOICE_ID_05685.wav|…… ʧo↓Qto, n e. fU↑ko↓ona ji↓koga a↓Qtano.
+louise/VOICE_ID_05686.wav|ku↓Q……!
+louise/VOICE_ID_05687.wav|to↑koro↓de mo↑NmoraN↓sii? a↑rata↓mete ta↑zuneta↓i ko↑to↓ga a↓ru N↓dakedo.
+louise/VOICE_ID_05688.wav|…… na↓Nde me↑seNo so↑ra↓su no↑kaʃira.
+louise/VOICE_ID_05689.wav|ʧa↑Nto da↑ma↓Qte te↓Qte i↑Qta↓deʃo! na↓Nde ko↑Nnani ba↑remakuQte↓runoyo!
+louise/VOICE_ID_05690.wav|ʦu↓i, j a na↓iwayo, mo↓o. hi↑miʦuQte i↑Qta↓noni ko↑reja, sa↑kiga o↑moiya↓rareruwa.
+louise/VOICE_ID_05691.wav|mo↑NmoraN↓sii. wa↑taʃi, do↑ojoo su↑ru↓wa.
+louise/VOICE_ID_05692.wav|so↑one. ja↓a, wa↑taʃiwa ju↓gyooga a↓rukara, o↑waQta↓ra ze↑NiNde ʃu↑ugooQte ko↑to↓de i↓i?
+louise/VOICE_ID_05693.wav|sa↓te, mi↑Nna a↑ʦuma↓Qtawane.
+louise/VOICE_ID_05694.wav|ʃI↑katana↓iwayo. mo↑NmoraN↓siiga ba↑ra↓ʃIʧaQta N↓da mo↑no. ko↑no sa↓i, te↓o ka↑ʃIte mo↑rau↓wa.
+louise/VOICE_ID_05695.wav|a↑ra, da↑ijo↓obuyo. ko↑no mu↑sume↓niwa, su↓guni de↑te Q↓te mo↑rau↓kara.
+louise/VOICE_ID_05696.wav|na↓niyo.
+louise/VOICE_ID_05697.wav|na↓niga hi↑do↓inoyo. wa↑taʃiwa a↑tarimaeno ko↑to↓o i↑Qte↓rudakeyo.
+louise/VOICE_ID_05698.wav|na↓NtokaQte, na↓niyo.
+louise/VOICE_ID_05699.wav|he↑emiN hI↑to↓rio wa↑taʃItaʧidakede ka↑kuma↓eQte i↑u↓no? so↑Nna n o mu↓ridawa.
+louise/VOICE_ID_05700.wav|mo↑tomotowa, ko↑no mu↑sume↓ga mi↑ʧibatade ta↑ore↓tetakara ka↓ihoo su↑ru ta↑me↓ni ga↑kuiNma↓de ʦu↑rete ki↓ta wa↓kedeʃo.
+louise/VOICE_ID_05701.wav|na↓ra bu↑gaiʃaga ga↑kuiNni i↑ru no↑wa o↑kaʃi↓ino.
+louise/VOICE_ID_05702.wav|wa↑taʃi, na↓nika ma↑ʧiga↓Qta ko↑to i↑Qte↓ru?
+louise/VOICE_ID_05703.wav|so↑reja↓a, wa↑taʃi↓taʧide ko↑no sa↑ki do↓o ya↓Qte ka↓nojono me↑Ndo↓oo mi↓te ku↓no?
+louise/VOICE_ID_05704.wav|i↑nu↓ya ne↓kojanainoyo. so↑Nna ko↑to, mu↓rini ki↑maQte↓rujanai.
+louise/VOICE_ID_05705.wav|w a?
+louise/VOICE_ID_05706.wav|n a…… Q! s o, s o, so↑Nna N↓janaiwayo!
+louise/VOICE_ID_05707.wav|so↑Nna ha↑zu na↓i! na↓Nde, ko↑no sa↑itoni ʃi↑Qtona↓Nte, ʃi↑nakUʧa na↑ra↓nainoyo!
+louise/VOICE_ID_05708.wav|u…….
+louise/VOICE_ID_05709.wav|wa↑ka↓Qtawayo! to↑obuNno a↑ida ka↑kuma↓eba i↓i N↓deʃo!
+louise/VOICE_ID_05710.wav|o↑so↓Qte ki↓ta re↑Nʧuuga, ji↑ʦu↓wa ka↓nojoo ʃo↑okaN ʃI↑ta ma↑hooʦu↓kaika so↑no bu↓kaQte ko↑to?
+louise/VOICE_ID_05711.wav|a, so↑ona↓no?
+louise/VOICE_ID_05712.wav|sa↑ito? a↓Nta, ko↑Ne↓Qʧina ko↑to ka↑Nga↓etadeʃo.
+louise/VOICE_ID_05713.wav|da↑Nge↓Nwa de↑ki↓naikedo, so↑oyuu ko↑to↓ni na↓ruwane.
+louise/VOICE_ID_05714.wav|…….
+louise/VOICE_ID_05715.wav|mo↓o o↑kIte↓ruwayo.
+louise/VOICE_ID_05716.wav|ma↑Qtaku, go↑ʃu↓jiNsamao ho↑Qtarakaʃini ʃI↑te gu↓uguu ne↑te↓runaNte, i↑Qtai na↑nisamano ʦu↑morikaʃira?
+louise/VOICE_ID_05717.wav|u, u↑rusa↓iwane. ho↑Nto↓owa ko↑nokuraini o↑kIte↓runoyo.
+louise/VOICE_ID_05718.wav|so↓Q, so↑Nna wa↑ke↓naideʃo. ʧa↑Nto ne↓tawayo.
+louise/VOICE_ID_05719.wav|da↓kara, ʧa↑Nto ne↑te↓ruQte i↑Qte↓rujanai! so↑reni i↓ʦumo ne↑te↓ru no↑wa a↓Ntadeʃo?
+louise/VOICE_ID_05720.wav|fu↑N, ya↑rukiga ta↑rinai↓noyo.
+louise/VOICE_ID_05721.wav|u…… ʧ i, ʧi↑gau↓wayo. he↑ya↓no ji↑Nkoomi↓ʦudoga ta↑ka↓ikara, ʧo↓Qto ne↑guruʃi↓kaQtadakeyo.
+louise/VOICE_ID_05722.wav|so↑reQte do↓oyuu i↓miyo!
+louise/VOICE_ID_05723.wav|ho↑Ntoona↓no? ma↓a, so↑rena↓ra i↓iNdakedone.
+louise/VOICE_ID_05724.wav|e? na↓Nde?
+louise/VOICE_ID_05725.wav|na↓Nde, wa↑taʃiga ʦU↑kai ma↑no ko↑odooo ʧI↑ku↓ji ka↑NʃI ʃi↑nakUʧa da↑me↓nanoyo! da↑i↓iʧi, sa↑itowa na↓nikao su↑ru ʦu↑moridaQta↓no?
+louise/VOICE_ID_05726.wav|so↑o? na↑ra↓iikedo.
+louise/VOICE_ID_05727.wav|mo↓ʃi, ʧa↑Q ka↓io da↓ʃIte i↑ta no↑nara, so↑reso↓ooono ba↓ʦuo u↑ke↓te mo↑rau↓kedone.
+louise/VOICE_ID_05728.wav|fu↓fuN, wa↑taʃiyo↓ri yo↑Qpodo o↑ne↓boosaNga i↑tamita↓ine.
+louise/VOICE_ID_05729.wav|ʧ o, ʧo↓Qto na↓ni a↑yamaQte↓runoyo! ʃi↑Qka↓ri ʃi↑nasaiyo.
+louise/VOICE_ID_05730.wav|so↑reyo↓ri, so↓rosoro ju↓gyooni i↑kanaito i↑kenai ji↑kaNda↓wa.
+louise/VOICE_ID_05731.wav|i↓ikara! sa↓Qsato ju↓gyooni i↑ku↓wayo.
+louise/VOICE_ID_05732.wav|a……!! su↓gu ki↑gae su↑ru↓kara, he↑ya↓kara de↑teQte!
+louise/VOICE_ID_05733.wav|ha↓yakU he↑ya↓kara de↑teQte! ki↑gaewa ʃi↑e↓sUtani ʃI↑te mo↑rau↓kara.
+louise/VOICE_ID_05734.wav|gu↓zuguzu ʃI↑tena↓ide, sa↓Qsato de↑teQte!
+louise/VOICE_ID_05735.wav|mo↓o, sa↑itoga ha↓yakU he↑ya↓kara de↑te ku↑renakaQta↓kara o↑kureta N↓deʃo?
+louise/VOICE_ID_05736.wav|e…… e↓e!? a, i↓e, so↑no, na↓Ndemo a↑rimase↓N. mi↓sUtako↑rube↓eru.
+louise/VOICE_ID_05737.wav|ha↓i.
+louise/VOICE_ID_05738.wav|m a, wa↑taʃi↓taʧiniwa ka↑Nkee na↓ikedone.

filelists/trainnn.txt

@@ -0,0 +1,691 @@
+wav/nen314_035.wav|0|まさかっ、契約の代償で？ぬいぐるみになるなんて、ありえるんですか？
+wav/nen314_036.wav|0|……見当外れだったんでしょうか？
+wav/nen314_037.wav|0|そのつもりです
+wav/nen314_038.wav|0|椎葉さん達も巻き込んで申し訳ありません
+wav/nen314_039.wav|0|でも、部活にもしばらく顔を出せなくなると思いますから
+wav/nen315_001.wav|0|犯人捜しに協力したい、ということですか？
+wav/nen315_002.wav|0|え、ええ
+wav/nen315_003.wav|0|いえ、保科君だとはわかっているんですけど
+wav/nen315_004.wav|0|ぬいぐるみとお話ししているようで、少し恥ずかしくて
+wav/nen315_005.wav|0|とにかく提案については、私としては是非お願いしたいというところなんですが
+wav/nen315_006.wav|0|七緒は可能性の問題だと言っています
+wav/nen315_007.wav|0|万一、犯人の可能性がある者へ情報が漏れて被害が拡大したら？そのリスクは背負えないと
+wav/nen315_008.wav|0|おそらくは
+wav/nen315_009.wav|0|ええ……昨日も言いましたが、私は犯人が許せません
+wav/nen315_010.wav|0|厚真さんは親しい友人の死を受け入れ、新しい友人との関係を築こうと必死でした……
+wav/nen315_011.wav|0|私が集めた欠片を吸収させて、なんとか安定はさせましたが
+wav/nen315_012.wav|0|知っての通り、これは一時しのぎに過ぎません
+wav/nen315_013.wav|0|はい、ですがこの場合、欠片は犯人から奪い返せばいいんです
+wav/nen315_014.wav|0|私の銃か、椎葉さんのハンマーで叩けばいいようです
+wav/nen315_015.wav|0|まだ私も試していませんが
+wav/nen315_016.wav|0|七緒によると、無理矢理削り取った欠片は元の持ち主へ返ろうとする力が強いようなんです
+wav/nen315_017.wav|0|だから散らしてしまえば、あとは自動的に戻るはずだと
+wav/nen315_018.wav|0|もちろん、そうなります
+wav/nen315_019.wav|0|ですから、手分けして捜査するというのはどうでしょう？
+wav/nen315_020.wav|0|一緒に行動するのは七緒の方針に逆らうことになってしまいますが、それなら問題ないはずです
+wav/nen315_021.wav|0|ええ、その上で私達以外の魔女やアルプがいないか捜索していきましょう
+wav/nen315_022.wav|0|それにまだ発見されていない被害者もいるかもしれません
+wav/nen315_023.wav|0|欠片を回収可能な人を見つけ出すのが難しいように、心が削れるほど弱ってる人を捜すのも難しいはずです
+wav/nen315_024.wav|0|見つけ出すだけでも、なかなか骨が折れそうですが
+wav/nen315_025.wav|0|もし……被害者が集中する地域が発見された場合、犯人のテリトリーを特定できるかもしれません
+wav/nen315_026.wav|0|そういうことです
+wav/nen315_027.wav|0|すでに七緒は周辺のアルプにも連絡を入れて、情報を募っているようです
+wav/nen315_028.wav|0|常習なら、他の地域でもやってる可能性がありますから
+wav/nen315_029.wav|0|ええ、私もまずは学院周辺から洗っていくつもりです
+wav/nen315_030.wav|0|椎葉さん達は駅の周辺から、捜査してもらえませんか？
+wav/nen315_031.wav|0|厚真さんのうちは駅の近くなんだそうです
+wav/nen315_032.wav|0|ええ、私からも報告は入れるようにします
+wav/nen316_001.wav|0|そうですか、そちらも手がかりはなしということですね……
+wav/nen316_002.wav|0|ええ……七緒も不思議がってるくらいで
+wav/nen316_003.wav|0|別のアルプがいるなら、匂いでわかるというのですが
+wav/nen316_004.wav|0|ともかく、簡単にはいかないということでしょう
+wav/nen316_005.wav|0|私が教えて欲しいくらいですから
+wav/nen316_006.wav|0|なにか思いついたら、相談させてもらいますね
+wav/nen316_007.wav|0|ええ、明日からは週末になりますし、携帯で報告し合うようにしましょう
+wav/nen317_001.wav|0|七緒がたまたま路上で倒れているところを発見したときには、すでにこの状態だったそうです
+wav/nen317_002.wav|0|ええ、もちろんです……七緒も、かまいませんね？
+wav/nen317_003.wav|0|……七緒？
+wav/nen317_004.wav|0|また椎葉さんを見張っていたのでしょうか？
+wav/nen317_005.wav|0|少なくとも、犯行を目撃した可能性は高いのでは？
+wav/nen317_006.wav|0|その道へ入っていったなら、保科君のお父さんが倒れていた場所は通り道のはずです
+wav/nen317_007.wav|0|その可能性は高そうですが……子犬？昨日、拾ったばかりなんですか
+wav/nen317_008.wav|0|いえ、厚真さんが預かっていた子犬も、行方がわからなくなっているのを思い出したんですが
+wav/nen317_009.wav|0|七緒、なにか？
+wav/nen318_001.wav|0|こんな子犬が犯人だったとは、見かけによらないようですね
+wav/nen318_002.wav|0|頼みます、七緒
+wav/nen318_003.wav|0|そうですか、厄介ですね
+wav/nen318_004.wav|0|私達が追っていることにも、気付かれたはずですから
+wav/nen318_005.wav|0|今後はますます、捜しづらくなりそうですね
+wav/nen319_001.wav|0|ええ、聞こえていますよ
+wav/nen319_002.wav|0|ともかく、これでなにかあっても対応できるはずです
+wav/nen319_003.wav|0|保科君っ、油断しないでください
+wav/nen319_004.wav|0|相手は人化の術を覚えている可能性もあるんですよ？
+wav/nen319_005.wav|0|人間に見えても、ぼんやりしないでしっかり警戒を
+wav/nen319_006.wav|0|もっと捜索範囲を広げてみますか？
+wav/nen319_007.wav|0|なるほど、それが動物の習性というものかもしれませんね
+wav/nen319_008.wav|0|いえ、申し訳ないのですが、私達の方が少し疲れてきました
+wav/nen319_009.wav|0|１０分ほど通話を切らせていただきたいのですが。ねえ、七緒？
+wav/nen319_010.wav|0|もちろん椎葉さん達は、会話を続けても結構ですから
+wav/nen319_011.wav|0|わかっていただけて、なによりです
+wav/nen319_012.wav|0|では、ごゆっくり。周囲の警戒は、私達でしておきますので
+wav/nen401_001.wav|0|すみません……あれはあくまで、あの時限りのピンチヒッターですから
+wav/nen401_002.wav|0|そう言ってもらえるのは嬉しいんですが……本当にごめんなさい。やっぱり人前に出るのは苦手ですから
+wav/nen401_003.wav|0|はあ……わかりました
+wav/nen401_004.wav|0|……もう、本当に疲れました
+wav/nen401_005.wav|0|ありがとうございます、戸隠先輩
+wav/nen401_006.wav|0|ふー……ふー………………はぁ、美味しい
+wav/nen401_007.wav|0|褒めてもらえるのは嬉しいんですが……もう少しゆっくりさせて欲しいです
+wav/nen401_008.wav|0|それ自体は光栄なことで、嬉しく思っているんですが……なんだか精神的に疲れてしまって……
+wav/nen401_009.wav|0|そうだとありがたいんですが……
+wav/nen401_010.wav|0|……わかりました。すみませんが、今日はそうさせてもらいます
+wav/nen401_011.wav|0|……お言葉に甘えてもいいですか？
+wav/nen401_012.wav|0|それではお先に失礼します
+wav/nen402_001.wav|0|そうですね。正直ホッとしています。これでオカ研の活動に集中できますし
+wav/nen402_002.wav|0|あ、今日は仮屋さんが誰かを紹介しに来るんですよね？
+wav/nen402_003.wav|0|はい、鍵は開いていますよ
+wav/nen402_004.wav|0|どうぞ、おかけ下さい
+wav/nen402_005.wav|0|いらっしゃいませ
+wav/nen402_006.wav|0|はい、もちろん大丈夫です
+wav/nen402_007.wav|0|はい
+wav/nen402_008.wav|0|そこの席にどうぞ、座って下さい
+wav/nen402_009.wav|0|それで、早速本題なんですが、悩み事というのは？
+wav/nen402_010.wav|0|………？すみません、まだちょっと状況が掴みきれないので、もう少し詳しい話を順序立てて教えてもらえますか？
+wav/nen402_011.wav|0|でも、今まで一緒にいた分、物足りないということですね
+wav/nen402_012.wav|0|……少し時間をいただけますか？相手の方にも確認したいので
+wav/nen402_013.wav|0|そうですね……とにかく、相手の本心を確認するのが最優先事項でしょうか
+wav/nen402_014.wav|0|そうですね。必要なのは言葉を伝えるだけのメッセンジャーではなく、コーディネーターですからね
+wav/nen402_015.wav|0|できれば先輩にお願いしたいんですが
+wav/nen402_016.wav|0|相手のことを考えると、年上の女性の方が打ち明けやすいのではないかと思いまして
+wav/nen402_017.wav|0|私だって得意じゃありません。ただ、私と戸隠先輩でしたら、先輩の方が話しやすいと思うんです。学生会長もされていたわけですから
+wav/nen402_018.wav|0|おかえりなさい、２人とも
+wav/nen402_019.wav|0|厄介と言いますと？
+wav/nen402_020.wav|0|ちょっと思いつきませんね
+wav/nen402_021.wav|0|そうですね……今日はひとまず解散して、明日また話し合いましょうか。それまで各自で考えてみるということで
+wav/nen402_022.wav|0|それでは、昨日の件について話し合いたいのですが――
+wav/nen402_023.wav|0|何か、考えが浮かんだ人はいますか？
+wav/nen402_024.wav|0|はい、因幡さん
+wav/nen402_025.wav|0|確かにマネージャーになれば、一緒にいることはできますね
+wav/nen402_026.wav|0|幸い、肥後さんは柴君のプレイする姿を恰好いいと言っていましたしね
+wav/nen402_027.wav|0|わかりました。それでは、柴君の気持ちと共に、バスケ部に入部という返答をしてみましょう
+wav/nen402_028.wav|0|保科君と戸隠先輩は、柴君と話をしてもらえますか？
+wav/nen404_001.wav|0|保科君、どうかしたんですか？授業中に呆けていたみたいですが
+wav/nen404_002.wav|0|そうですか？
+wav/nen404_003.wav|0|もし何かあるなら休んでくれてもいいんですよ？
+wav/nen404_004.wav|0|私もバンドの誘いが多いときは少し早めに帰らせてもらったりしたんです。皆さんも、たまには休んでもらってもいいです
+wav/nen404_005.wav|0|はい、遠慮なく
+wav/nen404_006.wav|0|わかりました。それじゃあ保科君
+wav/nen404_007.wav|0|なんですか？
+wav/nen404_008.wav|0|はい。もちろん構いませんよ
+wav/nen404_009.wav|0|それじゃあ、みんなには私の方から連絡をしておきますね
+wav/nen404_010.wav|0|気にせず。今日はゆっくりと休んで下さい
+wav/nen404_011.wav|0|はい。よろしくおねがいします
+wav/nen404_012.wav|0|体調が悪いわけじゃないと言ってましたが、心ここに非ずといった感じで
+wav/nen404_013.wav|0|ち、違います、違いますよぅっ！保科君がオカ研に入部したのは、恋愛感情とか、そういうことではなくて
+wav/nen404_014.wav|0|私に対する罪悪感といいますか、義務感と言いますか……それはきっと同情に近い感情ですから……
+wav/nen404_015.wav|0|―――ッ
+wav/nen404_016.wav|0|そ、それは……い、言えません
+wav/nen404_017.wav|0|もし、それを口にするぐらいなら……私、ここから飛び降りますっ。あと、秘密を漏らさないように、保科君にも一緒に死んでもらわないと
+wav/nen404_018.wav|0|す、すみません……そうですよね
+wav/nen404_019.wav|0|先に飛び降りたら、保科君に一緒に死んでもらうことはできませんからね。順番をちゃんと考えないと、まずは保科君を先に――
+wav/nen404_020.wav|0|そうなんですっ。事情があるんです！人には言えない事情が……恥ずかしすぎます……
+wav/nen404_021.wav|0|本当に、怪しいことは何もありません。保科君はいいお友達です。ただそれだけですから！
+wav/nen404_022.wav|0|それに保科君にとって、私はそんな対象じゃありませんよ。きっと……私のことなんて、ヘンタイだと思っているんですよぅ……ぅぅぅ……
+wav/nen404_023.wav|0|本当にそんなことはありません。部員が私一人でしたから。色々助けてもらっているだけですよ
+wav/nen404_024.wav|0|私も、思い当たりませんね。ただ、この中で言うのなら、私よりも戸隠先輩に惹かれるんじゃないでしょうか？
+wav/nen405_001.wav|0|ちょっといいですか？
+wav/nen405_002.wav|0|保科君。ああいうのは、どうかと思います
+wav/nen405_003.wav|0|ですから、戸隠先輩と学院内で……キス、とかしちゃうことです
+wav/nen405_004.wav|0|偶然通りかかって……。まあ、私の他には誰にも見られていないみたいでしたが
+wav/nen405_005.wav|0|学院内で、人気がないとはいえ、あんなところで……しかも、舌までレチョレチョと激しく……
+wav/nen405_006.wav|0|みっ、見えちゃっただけですよぅ！とにかく、ああいうのはいけないと思います。先生方に見つかったら事ですよ？
+wav/nen405_007.wav|0|もっとしっかりして下さい。人前であんなことしないように
+wav/nen405_008.wav|0|別に、迷惑だとかそういうことはないんですが……ああいうのを不意に見せられると、困るんです
+wav/nen405_009.wav|0|もちろんそういうこともありますが……その、あんなに激しいのを見せられてしまうと、私としてもドキドキしてきてですね……
+wav/nen405_010.wav|0|は……発情をしてしまう可能性と言いますか、実際に発情をしてしまったといいますか……――い、いえ！やっぱり何でもありません
+wav/nen405_011.wav|0|と、とにかく、私の体調のことは、この際無視をしてもらっても構いません。そうではなく、学院内ではもう少し落ち着くべきです
+wav/nen405_012.wav|0|線引きは、しっかりしないといけません
+wav/nen405_013.wav|0|はい、何ですか？
+wav/nen405_014.wav|0|どうと言われますと？
+wav/nen405_015.wav|0|そう言われると……そうですね。状況から考えると、戻っても不思議はありませんね
+wav/nen405_016.wav|0|今朝見たときには、それほど変化はなかったと思いますが……
+wav/nen405_017.wav|0|やっぱり大きな変化は……。一応、七緒のところで確認をしてもらいますか？
+wav/nen405_018.wav|0|そういう日もありますよ
+wav/nen405_019.wav|0|そうですね
+wav/nen405_020.wav|0|私はすぐそこで別れることになりますが、そこまででよければ
+wav/nen405_021.wav|0|お２人は、先に帰っていただいて大丈夫ですよ？
+wav/nen405_022.wav|0|私たちのことなら、気にしないで下さい。後片付けと言っても、施錠の確認をする程度の話ですから
+wav/nen405_023.wav|0|私でも気付くレベルでした
+wav/nen405_024.wav|0|いえ、今日は仕方ありませんよ。相談だけじゃなく、占いを希望する人も来ませんでしたからね
+wav/nen405_025.wav|0|はい、さようなら
+wav/nen406_001.wav|0|おはようございます、保科君
+wav/nen406_002.wav|0|後で少し時間をもらえますか？そんなにかかりませんから、お昼ご飯を食べた後で結構です
+wav/nen406_003.wav|0|はい、よろしくお願いします
+wav/nen406_004.wav|0|すみません。お呼び立てして
+wav/nen406_005.wav|0|そうですね。まず、これを見てもらえますか？
+wav/nen406_006.wav|0|はい。保科君に言われて確認をしたんです。そしたら少しずつですが、瓶の欠片の量が増えていました
+wav/nen406_007.wav|0|でも最近、私は回収を行っていないんです
+wav/nen406_008.wav|0|保科君の心の穴が埋まり、埋まった分だけ、この瓶に返ってきているんだと思います
+wav/nen406_009.wav|0|はい。それはいい事なんですが……気になるのはその量なんです
+wav/nen406_010.wav|0|欠片が戻ってきたのは、保科君が戸隠先輩とお付き合いをするようになったからだと思うんです
+wav/nen406_011.wav|0|正直なところ、そのわりには量が少なすぎる気がします
+wav/nen406_012.wav|0|確かに返ってきているのは間違いないんですが、最初は私も気付かないぐらいですから
+wav/nen406_013.wav|0|そこです。私が気になっているのは
+wav/nen406_014.wav|0|私が端から見てるだけでも、お２人はとても幸せそうに見えます。それなのに、気付けないぐらいの少量……ハッキリ言って変だと思います
+wav/nen406_015.wav|0|そうじゃないかと私は思っています
+wav/nen406_016.wav|0|何か後ろめたさを感じているようなことはありませんか？例えば……浮気をして、先輩を裏切っていることに罪悪感があるとか？
+wav/nen406_017.wav|0|ですよね。保科君がそんな人だとは思っていませんが……でもそれなら本当にどうして？何かしらの理由があると思うんですが……
+wav/nen406_018.wav|0|何か心当たりがありましたか？
+wav/nen406_019.wav|0|自分だけ隠し事をしていることに、罪悪感を覚えてるんですか……可能性は十分にありますね
+wav/nen406_020.wav|0|元々保科君の心の穴はその能力に起因しています。能力に負い目を感じている以上、穴が完全に埋まることはないんじゃないかと……
+wav/nen406_021.wav|0|それに……これはあくまで、責めるつもりではなく、色んな人の相談を受けて思った個人的な意見なんですが
+wav/nen406_022.wav|0|そういう負い目を放置してしまうと、いつか戸隠先輩との間に、わだかまりを作ることになると思います
+wav/nen406_023.wav|0|最近のお２人はとても幸せそうです。見ているだけで、それはわかります
+wav/nen406_024.wav|0|願わくは、このまま幸せでいて欲しいです。でもそのためには……今のままではダメなんじゃないかと思って……すみません。勝手なことを言って
+wav/nen406_025.wav|0|私にできることなんて、ほとんどありませんから。どうするかは保科君次第です
+wav/nen406_026.wav|0|はい、それがいいと思います
+wav/nen407_001.wav|0|もう夕方ですね、そろそろ終わりましょうか
+wav/nen407_002.wav|0|人と会う約束があると聞きましたよ
+wav/nen407_003.wav|0|それじゃあ、みんなで片付けましょうか
+wav/nen409_001.wav|0|保科君、気持ちはわかりますが、あまり無理はしないように気をつけて下さいね
+wav/nen409_002.wav|0|保科君まで倒れたら、きっと戸隠先輩も気にすると思います
+wav/nen409_003.wav|0|あ、保科君
+wav/nen409_004.wav|0|今日、帰りに七緒のところに寄ろうと思うんですが、なにか伝えておくことはありますか？
+wav/nen409_005.wav|0|はい。わかりました
+wav/nen409_006.wav|0|はい、また明日
+wav/nen409_007.wav|0|保科君、ちょっと
+wav/nen409_008.wav|0|今日、部活を休んで、戸隠先輩のお見舞いに行っても大丈夫でしょうか？
+wav/nen409_009.wav|0|七緒も自信はなくて、成功するかどうかもわからないんですが……一つだけ、試してみたいことがあるみたいです
+wav/nen409_010.wav|0|私や椎葉さんが欠片を回収するのと同じです。あれは具体的には、相手に魔法を打ち込み、気持ちを“欠片”という状態にして回収を行っています
+wav/nen409_011.wav|0|いえ、封印する魔法は常に発動しているので、外から解除するのは難しいらしくて
+wav/nen409_012.wav|0|そうして戸隠先輩の魔法に強い魔力で干渉して、不具合を引き起こさせて、封印している魔法を壊します
+wav/nen409_013.wav|0|そこに関しては大丈夫だろうって言ってました
+wav/nen409_014.wav|0|魔力の塊をぶつけることで、多少のショックを与えるかもしれないそうですが、先輩の心にひどい影響を与えるものじゃないそうです
+wav/nen409_015.wav|0|いいんです。保科君や戸隠先輩にはお世話になりましたから
+wav/nen409_016.wav|0|そうです。それにこのままじゃ、人の相談を聞く余裕なんてありません
+wav/nen409_017.wav|0|戸隠先輩を心配しているのは、保科君だけじゃないんですよ？
+wav/nen409_018.wav|0|そんな、頭を上げて下さい！
+wav/nen409_019.wav|0|言ったはずですよ？心配してるのは保科君だけじゃないんです。だから、お礼は必要ありません
+wav/nen409_020.wav|0|あと、七緒も来ますよ。駅前で待ち合わせです
+wav/nen409_021.wav|0|どうですか、七緒。戸隠先輩の様子は
+wav/nen409_022.wav|0|それじゃあ、やってみましょうか
+wav/nen409_023.wav|0|はい
+wav/nen409_024.wav|0|それで、具体的にはどうすればいいんですか？
+wav/nen409_025.wav|0|私は、この弾丸を撃てばいいわけですね
+wav/nen409_026.wav|0|私たちの？
+wav/nen409_027.wav|0|……わかりました
+wav/nen409_028.wav|0|それじゃあ一緒に、タイミングを合わせて
+wav/nen409_029.wav|0|戸隠先輩……起きて下さい。みんな、心配していますよ。もちろん、私たちもです
+wav/nen409_030.wav|0|部室で、待っています。だから……
+wav/nen409_031.wav|0|――戸隠先輩
+wav/nen409_032.wav|0|ですが七緒、今の反応を見ると、何の意味がなかったとも思えません。もう少し続けてみれば
+wav/nen409_033.wav|0|魔力のことなら、私は気にしませんから
+wav/nen409_034.wav|0|そんな……
+wav/nen409_035.wav|0|眠るのにも体力は必要ですから、快復しているということじゃないでしょうか
+wav/nen409_036.wav|0|いえ、別にそういうことはありませんが
+wav/nen409_037.wav|0|私も、できることがあれば手伝いますから、いつでも連絡して下さい
+wav/nen409_038.wav|0|それでは
+wav/nen409_039.wav|0|戸隠先輩、先輩がいてくれないと、悲しいです。また、一緒に誰かの悩みを解決させましょう。コーヒーを淹れて、待っていますから
+wav/nen410_001.wav|0|え？え？あ、あの……なにが……でしょうか？
+wav/nen410_002.wav|0|あ、そのことですか。改まって、何かと思いました
+wav/nen410_003.wav|0|怒ったら魔力を返してくれます？
+wav/nen410_004.wav|0|ふふ、すみません。今のはかなり意地悪な質問でしたね。ただの冗談ですよ
+wav/nen410_005.wav|0|怒ったりしてません。私も、戸隠先輩が目覚めてくれて安心できましたから。これぐらいですんだのなら、何の問題もありません
+wav/nen410_006.wav|0|それに、ほら。これを見て下さい
+wav/nen410_007.wav|0|戸隠先輩とのことがあったからだと思うんですが、欠片が戻ってきたんです
+wav/nen410_008.wav|0|保科君は全部を使ったわけじゃなく、一部を魔力として使っただけで、残りはこっちに戻ってきただけです
+wav/nen410_009.wav|0|戸隠先輩にもう負い目を感じることもなくなったからで、私のことを気にする必要はないんですよ
+wav/nen410_010.wav|0|それは、保科君がオカ研で頑張ってくれた分で相殺です。実際、今のこの欠片の量は、私が保科君と出会う前より、ほんの少し少ないだけですから
+wav/nen410_011.wav|0|ですから本当に気にしないで下さい。色々ありましたが……この際、なんやかんやで差引チャラということでどうでしょう？
+wav/nen410_012.wav|0|はい。よろしくお願いします
+wav/nen410_013.wav|0|おかえりなさい！
+wav/nen410_014.wav|0|そうでしたね
+wav/nen410_015.wav|0|実は、快気祝いと言うことで、ケーキを準備しているんです
+wav/nen410_016.wav|0|久島先生にお願いして、冷蔵庫で冷やしてもらっていたんです
+wav/nen410_017.wav|0|戸隠先輩は、好きなケーキはありますか？
+wav/nen410_018.wav|0|保科君は、どうしますか？
+wav/nen410_019.wav|0|私はフルーツタルトを
+wav/nen410_020.wav|0|おめでとうございます！
+wav/nen410_021.wav|0|………
+wav/nen410_022.wav|0|学院内ではあれほどダメだって言ってるじゃないですか
+wav/nen410_023.wav|0|ここでキスなんてしないで下さい。したら怒りますよ？
+wav/nen410_024.wav|0|それならいいんですが
+wav/nen410_025.wav|0|そうですね。誰かに見つからないうちに……でも、味わって食べましょう
+wav/nen411_001.wav|0|改めて、おめでとうございます。戸隠先輩
+wav/nen411_002.wav|0|でも寂しくなっちゃうのは、因幡さんの言う通りですね
+wav/nen411_003.wav|0|私たちはまだ卒業しませんから
+wav/nen411_004.wav|0|名残惜しくはありますが、私たちはそろそろお暇しましょうか
+wav/nen411_005.wav|0|それでは戸隠先輩、私たちはこれで。また会いましょうね
+wav/nen500_001.wav|0|あ、はい。わかりました。練習頑張ってくださいね
+wav/nen500_002.wav|0|手伝えることがあったら、何でも言ってくださいね
+wav/nen504_001.wav|0|保科君、調子はどうですか？
+wav/nen504_002.wav|0|そうですか。それならいいんですが、顔色がよくないように見えたので
+wav/nen504_003.wav|0|普段に比べると、という程度ですが。大丈夫ですか？
+wav/nen504_004.wav|0|練習のし過ぎは、身体によくないと思います
+wav/nen504_005.wav|0|そうですか？それならいいんです。じゃあ、本番を期待してますね
+wav/nen504_006.wav|0|はい
+wav/nen505_001.wav|0|どうですか？着方で、どこか間違えてるところはないですか？
+wav/nen505_002.wav|0|そうですか、よかった
+wav/nen505_003.wav|0|はい。可愛いですよ
+wav/nen505_004.wav|0|あ、保科君はバンドの準備がありますよね？ここは私たちに任せて、行ってくれていいですよ
+wav/nen505_005.wav|0|さてと、そろそろ時間ですね
+wav/nen505_006.wav|0|……もうすぐ本番の時間ですよ？心を落ち着かせる時間はもうありませんが……
+wav/nen505_007.wav|0|……ドキドキ……ドキドキ……
+wav/nen505_008.wav|0|えっと……こ、ここは、励まし会とか開いた方がいいんでしょうか？
+wav/nen505_009.wav|0|分かりました。そうですね
+wav/nen505_010.wav|0|鍵はここにおいて置きますから。それじゃあ、また週明けに
+wav/nen507_001.wav|0|見て下さい、保科君。ちゃんと、心の欠片が返ってきましたよ
+wav/nen507_002.wav|0|はい。他に考えられません、よかった……本当によかった……
+wav/nen507_003.wav|0|いえ、迷惑だなんてそんな。保科君が入部してくれて、本当に助かりました。ありがとうございます
+wav/nen507_004.wav|0|それで……これからどうします？
+wav/nen507_005.wav|0|今後の部活です。欠片も返してもらいましたし、これは保科君の心が埋まったということです
+wav/nen507_006.wav|0|ですから、無理に部活に付き合っていただかなくても……
+wav/nen507_007.wav|0|あ、はい。私はもちろん構いませんが………………大丈夫ですか？
+wav/nen507_008.wav|0|仮屋さんです。心が埋まったのは、仮屋さんと上手くいっているからですよね？
+wav/nen507_009.wav|0|なのに、部活を続けたりしたら、擦れ違いですとか、そういうことが心配になって
+wav/nen507_010.wav|0|埋め合わせ？
+nyaru/255.wav|2|きた
+nyaru/232.wav|2|結婚の人生の一つではあるけろ結婚以外のさ死事とらっぱにやりたいことだだった
+nyaru/330.wav|2|え電来が耐えルしかないない
+nyaru/370.wav|2|ウィドヤルネジョーヴィレアウロックされたさかつくってホル進んだよ
+nyaru/250.wav|2|入おし合せをったりは先生ま
+nyaru/302.wav|2|ラーシャンよ恋衛碑が今日の満連帯で私つ切婚してくさシンシラーシャン業
+nyaru/185.wav|2|八島シパターたりがとう
+nyaru/313.wav|2|テンコロマーレーにのかとわ
+nyaru/15.wav|2|暑れたのねちょっと見て行ぎたいと映いすあまじは皆さん
+nyaru/202.wav|2|イザやみたなとこらないと思んだけのう
+nyaru/291.wav|2|僕和アオレは糸がベイチアワーと結婚したえがらい好きになりました
+nyaru/299.wav|2|………
+nyaru/63.wav|2|………
+nyaru/82.wav|2|トがすきい
+nyaru/257.wav|2|幸ワクしいねる
+nyaru/377.wav|2|の気のちあれ内にしてあげてください
+nyaru/336.wav|2|待たな
+nyaru/396.wav|2|女の子と話すこともできがいし食事に刺うこともできません私は物質トではなく
+nyaru/197.wav|2|気持に義もな中そんなジビができてナロに長んとか
+nyaru/118.wav|2|イ岸すぐ一番木なる女のかにキアキや握りまさい
+nyaru/214.wav|2|電モをこれから社会時なって仕事すれてなっぱらいララ小活産興だ
+nyaru/173.wav|2|特たちが中学校下突き合い始めて
+nyaru/208.wav|2|大学生か
+nyaru/237.wav|2|い
+nyaru/282.wav|2|さな
+nyaru/58.wav|2|オケモナッピーアーええ在配神ためムれハイけフニー住枚
+nyaru/22.wav|2|軒変えてみぬ
+nyaru/120.wav|2|キラれトにこかわってしちゃいなさい
+nyaru/57.wav|2|クを懐つけ
+nyaru/111.wav|2|ターター今月も八月やったにま倒
+nyaru/8.wav|2|さ
+nyaru/198.wav|2|茶とでも疑問があるチはーあの結婚しなくて頭をあ布村な簡 単
+nyaru/400.wav|2|せんどうしたらいでしょうか
+nyaru/158.wav|2|い九に
+nyaru/125.wav|2|や漢ゼマレノアの考えだけのう
+nyaru/116.wav|2|あたーアータネーも八八八八八かって家なにへ
+nyaru/187.wav|2|よくあレ男女の気持ちのボルテージ
+nyaru/243.wav|2|のれ何をしての今ョアあのタ川カートやことでねやいましのり寝い好き
+nyaru/295.wav|2|知鈴がやったことがどれなけひどいことかを知って彼女に謝って近づかないように
+nyaru/32.wav|2|ああせっかが帰えれ
+nyaru/260.wav|2|ミ
+nyaru/50.wav|2|乳ビー天歳綾がおれ棚わたちどまでいた邪だな
+nyaru/31.wav|2|カをだっばわいた
+nyaru/399.wav|2|は留スで無く家庭も結構誘服ですしかしどうしても好きの子ができま
+nyaru/355.wav|2|兄か楽しいことは死味なっぱりとかに存んろに時間は使って
+nyaru/62.wav|2|ャンダテルからねま姉は淡雑素かエミナ探絡いたー業は許介た
+nyaru/55.wav|2|井短索にお願い事書くのは著作記なジやズガコレフリーンサイトなはかいジいです
+nyaru/366.wav|2|食べた
+nyaru/166.wav|2|平ガ化アーテエズン突き合った子ができたときに賞め初で知進なぐんだを頼いた
+nyaru/300.wav|2|ティシみ方タステシいシラシャンでおわわれ渡大社背中にでくれませんか百
+nyaru/317.wav|2|キ歌譲ていたしいゴールジゃな社ネ五免の前ごめのさい小月加なり
+nyaru/264.wav|2|味がおまりわぱてな
+nyaru/342.wav|2|早薬く
+nyaru/362.wav|2|キシア下砂ャブあが着くっておろいしまはい七会社よ
+nyaru/78.wav|2|へ行ったす
+nyaru/147.wav|2|ディスーンのキャークター裏ない
+nyaru/215.wav|2|っぱに変わってくれとむしれないました
+nyaru/327.wav|2|か忘れるのウともうけの忘れかたってわかなイラ時間目
+nyaru/53.wav|2|エイねヨ国のそまたのさん
+nyaru/199.wav|2|に結婚んでするものじなしい
+nyaru/270.wav|2|い助況をのってとたは
+nyaru/357.wav|2|何へ
+nyaru/99.wav|2|レジへ僕は女の子の手裁を触れたことがない同程ですボクロー
+nyaru/137.wav|2|か漏が月きがどかかがるいし人
+nyaru/140.wav|2|枠別れからムジャのあい何モ手を取りのクランドア何もなかんでうなすさね
+nyaru/70.wav|2|ズバアはあいいスと拝ケマケトり反対抜せでなすのワーレンセが事件でわ
+nyaru/43.wav|2|てタタのー背駅を追ってきたがら
+nyaru/79.wav|2|営用を何茶のちゃありがとう
+nyaru/363.wav|2|買い
+nyaru/7.wav|2|ナダーニーソに作ったんだ様学
+nyaru/112.wav|2|冷なのが
+nyaru/18.wav|2|に引耐方たねアイケンとか棚話どにせなかったま
+nyaru/310.wav|2|ナイ側があれーるトリコイ合っチャーへディチョも入ってレジゃんンとこまてのコ枚よ見
+nyaru/142.wav|2|自上年犬もスキンはあるんねよ
+nyaru/19.wav|2|でしっまい
+nyaru/265.wav|2|デミーニストとは受性の権利を見げ3ンズ業道と多様性を疾風
+nyaru/13.wav|2|何ええき駅か
+nyaru/284.wav|2|ワーラそ声話す
+nyaru/181.wav|2|もまだねきていません
+nyaru/278.wav|2|うんこら減転さ
+nyaru/323.wav|2|店から陰ンじゃないのおとりが写免
+nyaru/332.wav|2|今耐えるときで少なえる
+nyaru/296.wav|2|しました
+nyaru/274.wav|2|マリはのしう直きの
+nyaru/42.wav|2|った待ちねあ
+nyaru/376.wav|2|に出会れってなかなかないと思う
+nyaru/196.wav|2|までもあちょっとでもあ
+nyaru/127.wav|2|つらして
+nyaru/283.wav|2|自分がしよせがらいんじゃ
+nyaru/194.wav|2|気持ちな大きさんにさがあれパターンの迷みかが
+nyaru/178.wav|2|なさいええた先頃箱先彼じな帰告したらあすが
+nyaru/48.wav|2|わたです田田か電話したから
+nyaru/66.wav|2|円
+nyaru/47.wav|2|だあ
+nyaru/285.wav|2|あれ自分なしャーせがらいいます
+nyaru/287.wav|2|に
+nyaru/220.wav|2|古的には朝仏婚絡け婚落な若い精神をげたのに切婚
+nyaru/121.wav|2|フラレルのをまた経験のかな
+nyaru/319.wav|2|番え声えのえ声はジスキカがとがいなさい五園なさい
+nyaru/381.wav|2|朝兄ャレちゃんわたしも二銭
+nyaru/252.wav|2|のミチ上も飛ぶな継飯らが読みな
+nyaru/188.wav|2|………
+nyaru/186.wav|2|難短しいねよ
+nyaru/122.wav|2|がてなてはどしたにウェチさーマーレ的アね礼愛しないかららんかよ
+nyaru/119.wav|2|アータン変コーダしないから彼者ができなえなん
+nyaru/403.wav|2|ない可能性はある
+nyaru/249.wav|2|やしいな真黙ってくれ
+nyaru/192.wav|2|つこまでん
+nyaru/160.wav|2|毎へちけってらね
+nyaru/246.wav|2|………
+nyaru/259.wav|2|天二女な卵声を踏みった渡し天船ミを使った天二人
+nyaru/110.wav|2|つません上車すればいなあったドがテ画をがっておいのはす
+nyaru/204.wav|2|商会一人の人人とを結婚してさが巻いけたらいちはいいたなさ
+nyaru/405.wav|2|えかけられる気がするんだねなす女な子白れないていなた勝手に話しかけられるいたいな
+nyaru/81.wav|2|タレダよペータールた四店ーマをレは
+nyaru/174.wav|2|いた
+nyaru/349.wav|2|忙しッくするかも考えら暇もないのらい
+nyaru/386.wav|2|ツ帯に関してはーマーレン対するなんかでい厭がらせがったにるマレがこる屋がが
+nyaru/353.wav|2|からないようにしたほうがあ辛らい時間はね
+nyaru/184.wav|2|ナレオホの年これはレア経験相談相談ま州がった
+nyaru/161.wav|2|好きや毎手を握ぎれのなく問題なきませね
+nyaru/340.wav|2|ま
+nyaru/226.wav|2|つそこいへは考内と一台に言えないけの
+nyaru/280.wav|2|なく何わー
+nyaru/347.wav|2|孫りきごしたすらに員
+nyaru/103.wav|2|次顔じゃなよ部んドワーティでらくず患じ側上なけてらったもチアチアエこチ
+nyaru/145.wav|2|………
+nyaru/67.wav|2|レガ本頼の髪の拝近です
+nyaru/222.wav|2|ン家園に思うきもあれかもしれないからさ
+nyaru/401.wav|2|これはねあの窓リがの君の顔を見てないから何とも言えないけた
+nyaru/322.wav|2|あ無ちろん自分なてしまった広いこといちゃんとなんかって謝ったり
+nyaru/86.wav|2|ァレアッタオリあれある中コクハクトカする日がの胸
+nyaru/167.wav|2|かったハリュアチチィ弱わたテンシと着がいなくなさい
+nyaru/14.wav|2|のでこやり愛しましたを板のむちょとむしろい経験だっぱりとか
+nyaru/239.wav|2|子ミ千マオレゃ自牧心が見るのが大好きみたことがいだけぬ
+nyaru/162.wav|2|いいん社会いでしゃんか
+nyaru/275.wav|2|リナ人それんぞれ
+nyaru/233.wav|2|目も人生なかなん
+nyaru/37.wav|2|今す
+nyaru/21.wav|2|駅帰えるえ
+nyaru/245.wav|2|の上気店
+nyaru/333.wav|2|これ生いでらえはコらイですかないぞいまわ
+nyaru/391.wav|2|で反聖してさかをつこってもたフロ死きてくさい
+nyaru/277.wav|2|まがどれがいいと学愛い水章はないな
+nyaru/367.wav|2|さ空げ厚くってへふルへジェェッたった
+nyaru/87.wav|2|木を汚らが着足ない大井ジャミ名マーレーの聞いていたが
+nyaru/352.wav|2|ーデッてル時間が一番か日に背れたのなかナメーカ観白難時下
+nyaru/324.wav|2|から許してくれいと思う
+nyaru/30.wav|2|つりは今気づいた
+nyaru/164.wav|2|その時にな中大をどしたり折レへ味てたいにぎるんだやメんトかラんかを言う
+nyaru/94.wav|2|残身な残心なあい置にしよう
+nyaru/152.wav|2|ままでもねあのはコ道な入り猫タ外ぬ悪い骨でもないからねあの反対楽しんで
+nyaru/183.wav|2|とぶ好きやれてくしゃりいんじゃったが
+nyaru/2.wav|2|木グラグラ毎診乱ですけろは
+nyaru/27.wav|2|なわたてん
+nyaru/150.wav|2|前突してな最興のキャーにしてね住年遊んだほうがいる
+nyaru/83.wav|2|店
+nyaru/354.wav|2|らセれと思わなで
+nyaru/397.wav|2|博自身はあれんたね私しを読室に何挙な利近円んてことか
+nyaru/5.wav|2|麻社をエたゴーダンシャー電チャわ三派を室声員
+nyaru/385.wav|2|た分自動でブロクされた話と思います
+nyaru/10.wav|2|ンはへ行た
+nyaru/224.wav|2|今
+nyaru/131.wav|2|思って好きアがあって思った人要がへできたときりいんじゃあい
+nyaru/247.wav|2|………
+nyaru/17.wav|2|ほしいねないようは入タテタラニを毛皿のタへ行好いたまたまつ
+nyaru/46.wav|2|アエャ道部じっと映ア皿たがあって
+nyaru/193.wav|2|木であら安好きなけどそを熱ってよ
+nyaru/98.wav|2|ねえね
+nyaru/104.wav|2|し案世らに馬鹿につげはいてこたわ
+nyaru/404.wav|2|トヤナム毎チ星月チの一きめなったがマりから
+nyaru/380.wav|2|切ちんありたあ
+nyaru/286.wav|2|投げらんあんまりあんか公がもコメントしなうせそ
+nyaru/134.wav|2|なのねえああ中反ン対好きの猫ができたご歌すねわ
+nyaru/382.wav|2|な中にブロックサれで泣きましたってごめんささいは
+nyaru/227.wav|2|よもま
+nyaru/231.wav|2|まとここに結婚察もう不面ではないけど文術やぎたかったこととかあったのになあっとか
+nyaru/157.wav|2|歳部屋にやったす
+nyaru/107.wav|2|すかま七とはでね
+nyaru/321.wav|2|コ寺院業派��引き出したカットティへのアーちょっと搬選しても神いけど
+nyaru/258.wav|2|ニむペミ二店ぬ
+nyaru/205.wav|2|たかもちく大事に結婚さっておいてもいいんじゃないでしゃょうか
+nyaru/314.wav|2|マ水ィナメメ学チャンエデチョンに売転ーレジャろ々ながってした
+nyaru/175.wav|2|十茶次ぎ
+nyaru/348.wav|2|本や活造したいいいい
+nyaru/369.wav|2|さた
+nyaru/190.wav|2|もちレシんのさいたいな
+nyaru/219.wav|2|ナーさん台は二十降半のかなってたなさあちょっと話を買ってくりよねさ女の
+nyaru/201.wav|2|真九人によるかもしてなきのさあア毎者にた立婚定ディコン川
+nyaru/29.wav|2|ケーとか帰るよかったが
+nyaru/326.wav|2|真類と立ついんだっぱな
+nyaru/293.wav|2|ため彼女にアプろうちしたり個人情報を引き出したかった
+nyaru/114.wav|2|うちらん地すか日本だよテージャファー
+nyaru/297.wav|2|でも今でもきどき胸が痛いですはきそうになります
+nyaru/3.wav|2|あノースさんでろす
+nyaru/305.wav|2|でこ孫レに対子行ってぬ上れん
+nyaru/371.wav|2|イナーレ車われだきじないち上って入神ージ店ラ員なのかパ
+nyaru/69.wav|2|堀バンは良しれたんなぽいすむ
+nyaru/172.wav|2|つもりなんですけのう
+nyaru/92.wav|2|みんなだ
+nyaru/200.wav|2|とあ
+nyaru/292.wav|2|でも彼女はがチコ以上要らないらしくてそれでもずっと自分の声が見る
+nyaru/169.wav|2|なげはランカを別ちで初や手能は観難かないとめま
+nyaru/25.wav|2|は一回少こままね
+nyaru/289.wav|2|えた
+nyaru/364.wav|2|店
+nyaru/281.wav|2|れないほがいいよね服し
+nyaru/244.wav|2|だう
+nyaru/80.wav|2|周りが欲しいダだヤ
+nyaru/141.wav|2|木が五るのレオさーバロラートムス近書ぎすればさー
+nyaru/23.wav|2|ア駅へ棚々にすれ玉たできるかな
+nyaru/115.wav|2|ーレーミあルジーピー員マーレーウールジアピーでスチャーピン
+nyaru/308.wav|2|モレにみんなエモレは逆リ逆院ー名ネあルジーモーレにホれれ前
+nyaru/165.wav|2|度出さいからなく知れっくつなげはいいとを
+nyaru/368.wav|2|………
+nyaru/212.wav|2|ルはね
+nyaru/88.wav|2|小がし会って持けた人嘘井じゃんき上ってゃよも
+nyaru/279.wav|2|どれが制ぎてったら誰かが敵になるからさ
+nyaru/254.wav|2|………
+nyaru/51.wav|2|とこリちゃんとこら
+nyaru/144.wav|2|若りはなあ
+nyaru/331.wav|2|歯です
+nyaru/339.wav|2|あのは
+nyaru/168.wav|2|テッチレットつないね
+nyaru/151.wav|2|やなやな
+nyaru/35.wav|2|わた棚音の背景あこれたかへいたな
+nyaru/217.wav|2|てもさ例上ケの社迦会神軍連メト館になっててえ
+nyaru/251.wav|2|線年は
+nyaru/267.wav|2|インガー野拝エアーチスス中にセミー室でること防ヒャンしたに
+nyaru/335.wav|2|うん
+nyaru/225.wav|2|あの君に使ってくれているのかな
+nyaru/393.wav|2|2時2歳難ですけのは女の子の手をちなぐことすらできませんでしたバーコリ行イな
+nyaru/341.wav|2|どうしての忘れたいなったら別のイさんを押しすっていのが一番なんかぺっとり
+nyaru/242.wav|2|七回し
+nyaru/28.wav|2|ない人転論か
+nyaru/389.wav|2|てくどさえ
+nyaru/0.wav|2|地学閉産機さ消すが
+nyaru/343.wav|2|バスに幸り変じゃが行いかぞ
+nyaru/398.wav|2|が玄学ことぼおどかとかな
+nyaru/146.wav|2|イーか原身に国き白いのジ現進見え
+nyaru/298.wav|2|どあしたは彼女を忘れられますか教えてください
+nyaru/108.wav|2|わざメンファータにする
+nyaru/138.wav|2|お金お高に使ったーはぎじゃないバロランとんンスキ買っパはぎじゃない
+nyaru/210.wav|2|来なきの結婚できをとした
+nyaru/124.wav|2|かったーや誤しませてはいい痛は
+nyaru/1.wav|2|裏に言を最初からげなげだランで好きのあ
+nyaru/52.wav|2|ちゃんントごらほら七たでよごらす
+nyaru/195.wav|2|ら目ざかしいよね
+nyaru/128.wav|2|洗い
+nyaru/209.wav|2|社会人家とは観来だけどは
+nyaru/373.wav|2|てて
+nyaru/365.wav|2|ですくった
+nyaru/71.wav|2|罪の電気飲まれぬ発はしか次恋ア教とりなってで
+nyaru/372.wav|2|黒やったくフな息度ねへやったこなきのをシアラネタラーソン技つるはもれ
+nyaru/303.wav|2|八し夜りがとうあ
+nyaru/266.wav|2|せれしたのことで
+nyaru/95.wav|2|右に
+nyaru/61.wav|2|気に捨てから我れ洗得眠道へてしらえてちゃんつえつてかす
+nyaru/346.wav|2|は神来や５年熱まぞレーラったらどう忘れるんだろうな
+nyaru/253.wav|2|立策便ンファータ
+nyaru/301.wav|2|善ィ体な配生を有れがテイトリ刺をくれウシた奴じゃー珍
+nyaru/213.wav|2|駒川生徒かなったら併ない結婚しって友をして大学生
+nyaru/73.wav|2|机クれ方はは電日��ます買く定オメナマ上や廃信で勉強セ関除つくろでして
+nyaru/315.wav|2|もえせせて何ピスのアヤりジ万れぬえぬ
+nyaru/290.wav|2|マれ
+nyaru/56.wav|2|ラフリーソザイです濃も電塞いたのは
+nyaru/75.wav|2|節つふったいよ
+nyaru/318.wav|2|わした土電犬さんスカった
+nyaru/191.wav|2|八かな結お彼女山のほが色しきて撮制君まだ
+nyaru/41.wav|2|ゲルトねこれをを考わた
+nyaru/189.wav|2|し町ちのラズメーターの高さが会わな買ったときの
+nyaru/240.wav|2|んはドブンはさすね
+nyaru/334.wav|2|はらい見ちかないすね
+nyaru/379.wav|2|はいつぎえええこちら兄ちゃんは夜波端
+nyaru/344.wav|2|思います
+nyaru/44.wav|2|書気なしてご歌円から零れ
+nyaru/262.wav|2|で意味が分かいち売天
+nyaru/234.wav|2|フンターに
+nyaru/395.wav|2|もちろんが八つこをしたこともハチこをしたこともなくてえ
+nyaru/387.wav|2|どもった朝伝倒したに何カも何カーをしてきた人アくしてると思いまい
+nyaru/156.wav|2|気れくなレター今日今から禁験すようね周りに決捨てきてルしているでにやった
+nyaru/59.wav|2|際となあ
+nyaru/216.wav|2|ねんちょっと
+nyaru/263.wav|2|ごしいこけにパねなぜ1世ランちャらカっちゃらだけ
+nyaru/351.wav|2|やば何もしない時間とか特にその寝れ前の勤んであねげなへら
+nyaru/229.wav|2|チェトチョットワラン員添息落っちじいよっての砂のに入ったほえいいん社けですょうか
+nyaru/40.wav|2|の
+nyaru/271.wav|2|２間でのデ番で見を乗ってい
+nyaru/102.wav|2|ぐむ感じへ上なかった馬をンすれはいてこたく
+nyaru/9.wav|2|は一日会戸猫ねですね
+nyaru/312.wav|2|マーレーではない６コマオレのこともしかせて
+nyaru/235.wav|2|わなのかってちゃんと考えてから結婚したほうがいいたお前ませ猫らわ
+nyaru/268.wav|2|プロった慈屋にきれる剣でストイしたいす現れてで少し文持知さいて
+nyaru/132.wav|2|かいない党を和いかのしれないみてんが
+nyaru/6.wav|2|願シんであげまちょうは
+nyaru/325.wav|2|またがの
+nyaru/171.wav|2|ええマーレン先生僕下の彼じは問すぐ日本な利学生活が終わって地惑に戻る
+nyaru/117.wav|2|彼女すらいないてくたわ
+nyaru/72.wav|2|年来三月に向けて患字で保し運画がマオ齢生制な患女
+nyaru/272.wav|2|なんだ
+nyaru/338.wav|2|の退決してくれるか
+nyaru/356.wav|2|行きましゃあ
+nyaru/149.wav|2|よとつしてね
+nyaru/218.wav|2|あても何年も食てて女なことから年も二ジある
+nyaru/307.wav|2|いった
+nyaru/16.wav|2|パナーバタャを見いとござましていのがわたっく何いね
+nyaru/241.wav|2|ニ業金があるからとことは
+nyaru/374.wav|2|二発クラッパナー赤ハーンパーにやでもそねて礼い掛けたげな好きな人ができ
+nyaru/170.wav|2|すぎえ
+nyaru/159.wav|2|握ったこつないのか
+nyaru/378.wav|2|………
+nyaru/84.wav|2|あ河が落ちるかなヴァレンタインゲは非常に漢情的ね馬
+nyaru/113.wav|2|エソセェクトッちなうあってね
+nyaru/236.wav|2|てての抜決願するな知にしないがなまねん
+nyaru/36.wav|2|ッパ開ちてか
+nyaru/261.wav|2|セミーミースとってことま
+nyaru/154.wav|2|かは椅しに大好ラああってもしたかできたら
+nyaru/105.wav|2|地の地て漢字はへタキで漢じがあいとあて
+nyaru/223.wav|2|取り隣会さの女ろ子の人生を
+nyaru/33.wav|2|大月がユンジゃラタイラジない植に
+nyaru/68.wav|2|そのバアの背駅与えれますか
+nyaru/129.wav|2|ツラー雨二さん
+nyaru/304.wav|2|屋食う
+nyaru/60.wav|2|空なので来蛇夫です野緒作金頭もりメ着きをしてるから大時
+nyaru/4.wav|2|今小ラシャ曜へとチキーンズチャン撮勢毎車ャ浴山半を
+nyaru/276.wav|2|いろんな考えがあれかや
+nyaru/97.wav|2|目ぬままろこちら円
+nyaru/130.wav|2|アヤテへそしずとしにるとかもよく若のいすかジ文アコン当に突きあいたいと
+nyaru/337.wav|2|時間とね新しい声が
+nyaru/392.wav|2|はいでこち洗しはり上せ洗ますええったじャちゃん私し
+nyaru/64.wav|2|当索じゃない起れちゃんと木が止れてます
+nyaru/38.wav|2|今たうよです
+nyaru/155.wav|2|あの絶対に後悔えないように当たっくしたわいと思わす
+nyaru/90.wav|2|った大丈が地ちんじゃ腹れにはねにかすねえ
+nyaru/133.wav|2|なだスれラか別に和いと思わなくいいと思い
+nyaru/256.wav|2|えれか
+nyaru/306.wav|2|マレがチ声でまいていとっけ
+nyaru/143.wav|2|っちがとくか
+nyaru/320.wav|2|大歳答う通りじガ部イは散かなすええったー
+nyaru/136.wav|2|や胸でして突きあほがついよちっ妬は
+nyaru/96.wav|2|いに甘字はです
+nyaru/402.wav|2|うもしかしたらオレいメ男全テオラが尋連に月チャって持て
+nyaru/388.wav|2|真中何ニモリ許しにブロックすることはないからもち自分の行能アの踏に返ってみ
+nyaru/329.wav|2|末たも時間か新らンしい今やあの傷を癒してくれると思うの
+nyaru/384.wav|2|また部エネジーマードついたかなんかれえ
+nyaru/345.wav|2|………
+nyaru/126.wav|2|ながイルジャーナかつの怪密が入る法だ依いみたいな資向の人デまわれ巡ゼソも
+nyaru/89.wav|2|マニアー今からランする救急を遅れ樹席木を取れ納ないち
+nyaru/269.wav|2|海撃つ
+nyaru/109.wav|2|う
+nyaru/45.wav|2|り年じゃない
+nyaru/311.wav|2|………
+nyaru/26.wav|2|電物で北わたりです
+nyaru/358.wav|2|ブラックリストに入ねだれたらどうしますか青これねえラック・礼へ押し
+nyaru/20.wav|2|駅へた丸まったとかのばよかったかな
+nyaru/179.wav|2|に結婚してほしいと寄れました
+nyaru/54.wav|2|な
+nyaru/163.wav|2|中ではさテービギッたことラいないんだきのサー初めて店レ根とき構うこの先くるとし
+nyaru/135.wav|2|今れなく別に
+nyaru/207.wav|2|会ってて流学生脱が終わって物ってくれなからあココーシェ加いま
+nyaru/85.wav|2|帯以社じゃ
+nyaru/74.wav|2|だやめたほがいい厭
+nyaru/230.wav|2|いった父インんじゃないでしょうかここへ流先恋士に今済みのまれな
+nyaru/39.wav|2|らぱ背かクラし皿を絶すねか急いた田が入きなす
+nyaru/273.wav|2|あかぶの七張い七中それずれのからに回りなこっぺスにでめれまいて思います普タた
+nyaru/350.wav|2|の愛しクしてま忘れるかもちれまい
+nyaru/211.wav|2|高校家大学かく社会人なへがとうんだけろう今だすぬ年教や
+nyaru/375.wav|2|ていいこだよね砂に看残ってもそ日キなレレし
+nyaru/182.wav|2|そんな僕がどうやって彼女と向い合へはいいですか
+nyaru/77.wav|2|ん
+nyaru/180.wav|2|かしにとっては結婚に対する心の準利でも生活に対する準備で
+nyaru/12.wav|2|マシマロが麦ィカの恋愛体験だった兄へ専愛に単時間する悩ミな
+nyaru/176.wav|2|あっちので
+nyaru/228.wav|2|だからといてラカさ土合場で結婚すれれお可哀相案がら破え
+nyaru/49.wav|2|棚はたれ
+nyaru/91.wav|2|早にやまいと
+nyaru/390.wav|2|い
+nyaru/316.wav|2|でね
+nyaru/24.wav|2|ちゃっとまってん
+nyaru/139.wav|2|ギャランテさ宅井はレトして漢ズにお金をきつかったどにしてさあ
+nyaru/328.wav|2|決連れた魔ジであのいろんな人がい色なところでやってるてもけどまチ時間が大決
+nyaru/238.wav|2|何女施送心うくい回り奥さんの
+nyaru/294.wav|2|して結果的に彼女をひどっくひどく傷つけてしがいました
+nyaru/360.wav|2|ラク内茶ャををっぱン電
+nyaru/123.wav|2|なとかなイン出別にあの恋愛ごり押しでゃないからさあ達ちに歯好きが人がいな
+nyaru/93.wav|2|のレがここにきて握に出すくもしまろ
+nyaru/34.wav|2|のワットで熱た棚っのはいてあるから花人が夜ラな花りじゃ員
+nyaru/248.wav|2|かしいなあ
+nyaru/394.wav|2|チオれと近いね海灰灰半員
+nyaru/148.wav|2|ち
+nyaru/101.wav|2|ノアンチィア裏かわかないだけのう合しをしてくださいぶ馬倒をすればいってことは
+nyaru/76.wav|2|な八た前しら自の折れたゃパーイン節たありがと
+nyaru/11.wav|2|の前へ同体で干シ案した電ア員
+nyaru/65.wav|2|ドリーそです
+nyaru/206.wav|2|今何歳なのかは科来だけどとこれ年市にも選なー地学校から
+nyaru/106.wav|2|かぜアタイトル互恋愛はニファータに住店
+nyaru/359.wav|2|話だけど地が好きな人に押し書かにブラックリッスいられたら幕が
+nyaru/288.wav|2|ちょっと対借の麻車まらあ兄がとうございます
+nyaru/153.wav|2|疑いに気ならくね
+nyaru/203.wav|2|ナレメックさーまに婚しないね
+nyaru/309.wav|2|画にと思ってにわは
+nyaru/383.wav|2|あええ中ブロックチョリビリイリノをかんンドブロッカ回レは一構したことないのねえ
+nyaru/100.wav|2|なてものはプレ犬超しをしてくださいのか立ちしだちん定
+nyaru/221.wav|2|してくれな店何にいたいま
+nyaru/361.wav|2|ーて気のを寒がつくってほんされタた
+nyaru/177.wav|2|季室すとおめん

filelists/trainnn.txt.cleaned

@@ -0,0 +1,410 @@
+wav/nen507_007.wav|0|a, ha↓i. wa↑taʃiwa mo↑ʧi↓roN ka↑maimase↓Nga.................. da↑ijo↓obudesUka?
+wav/nen507_008.wav|0|ka↑riyasaNde↓sU. ko↑ko↓roga u↑maQta no↑wa, ka↑riyasaNto u↑ma↓kuiQte i↑ru↓karadesUyone?
+wav/nen507_009.wav|0|na↓noni, bu↑katsuo tsu↑zuketa↓ri ʃI↑ta↓ra, su↑reʧigaide↓sUtoka, so↑oyuu ko↑to↓ga ʃi↑Npaini na↓Qte.
+wav/nen507_010.wav|0|u↑meawase?
+nyaru/255.wav|2|ki↓ta.
+nyaru/232.wav|2|ke↑QkoNno ji↓Nseeno hI↑to↓tsudewa a↑ruke↓ro ke↑QkoNigainosa↓ʃI ko↑to↓to ra↑Qpani ya↑rita↓i ko↑to↓dadaQta.
+nyaru/330.wav|2|e↑deN↓raiga ta↑eru↓ʃIka na↓inai.
+nyaru/370.wav|2|wi↓doyarunejoovireauroQkusaretasaka tsU↑ku↓Qte ho↓ru su↑suNdayo.
+nyaru/250.wav|2|nyu↑uoʃia↓waseoQtariwa se↑Nsee↓ma.
+nyaru/302.wav|2|ra↑a↓ʃaNyo ko↓i ma↑moru↓higa kyo↓ono ma↑Nre↓Ntaide wa↑taʃItsu se↓tsU ko↓N ʃI↑te ku↓sa ʃi↓N ʃi↑raaʃaN↓gyoo.
+nyaru/185.wav|2|ya↑jima ʃI↑pa↓taatariga to↓u.
+nyaru/313.wav|2|te↓Nkoromaareeni no↑katowa.
+nyaru/15.wav|2|a↓tsuretanone ʧo↓Qto mi↓te gyo↑ogitaito u↑tsu↓i su↑ama↓jiwa mi↑na↓saN.
+nyaru/202.wav|2|i↓za ya↑mita↓nato ko↑ra↓naito o↑moe↓Ndakenoo.
+nyaru/291.wav|2|bo↑kuwaaorewa i↓toga be↑iʧia↓waato ke↑QkoN ʃI↑ta e↑gara↓i sU↑ki↓ni na↑rima↓ʃIta.
+nyaru/299.wav|2|.........
+nyaru/63.wav|2|.........
+nyaru/82.wav|2|to↓ga sU↑kii.
+nyaru/257.wav|2|ko↑o↓wakU ʃi↑i↓neru.
+nyaru/377.wav|2|n o k i no↑ʧi a↑re↓naini ʃI↑te a↑ge↓te ku↑dasa↓i.
+nyaru/336.wav|2|ma↑te↓tana.
+nyaru/396.wav|2|o↑Nna↓nokoto ha↑na↓sU ko↑to↓mo de↑ki↓ga i↓ʃI ʃo↑kujini to↑ge↓u ko↑to↓mo de↑kimase↓N wa↑taʃiwa bu↑QʃItsutode↓wanaku.
+nyaru/197.wav|2|ki↑moʧini gi↓mona na↓ka so↑Nna ji↓biga de↑ki↓te na↓roni na↓ga N↓toka.
+nyaru/118.wav|2|i↓kIʃi su↓gu i↑ʧi↓baN ki↓naru o↑Nna↓no ka↑ni↓ki a↑kiya ni↑girima↓sai.
+nyaru/214.wav|2|de↓N mo↑o ko↑rekara ʃa↑kai↓ji na↓Qte ʃi↑goto su↑re↓te na↑Qpa↓rairara ʃo↑okatsu↓saN kyo↑oda.
+nyaru/173.wav|2|to↑ku↓taʧiga ʧu↑ugaQkoo↓ka tsU↑ki a↓ihajimete.
+nyaru/208.wav|2|da↑iga↓kUseeka.
+nyaru/237.wav|2|i.
+nyaru/282.wav|2|sa↑na.
+nyaru/58.wav|2|o↓ke m o na↑Qpi↓iaaee za↑ihai↓ʃiN ta↑memurehai↓ke fu↑niijuu↓mai.
+nyaru/22.wav|2|no↑kI ka↑ete mi↑nu.
+nyaru/120.wav|2|ki↓rare t o ni↑ko ka↑waQte ʃI↑ʧainasai.
+nyaru/57.wav|2|ku↑o na↑tsUketsUke.
+nyaru/111.wav|2|ta↓ataa ko↑Ngetsumo ha↑ʧigatsu ya↓Qta ni↑ma ta↑ose.
+nyaru/8.wav|2|s a.
+nyaru/198.wav|2|ʧa↑todemo gi↑moNga a↓ruʧiwaa a↑no ke↑QkoN ʃi↑nakUte a↑tama↓o a↑nunomurana ka↓N ta↓N.
+nyaru/400.wav|2|se↑Ndoo ʃ I ta↑raideʃoo↓ka.
+nyaru/158.wav|2|i kyu↓uni.
+nyaru/125.wav|2|y a ka↑Nzemare↓noano ka↑Ngaedakenoo.
+nyaru/116.wav|2|a↑ta↓a a↑ataneemo ha↑ʧima↓N ha↑Qse↓N ha↑Qpyaku ha↑ʧijuu ha↑ʧi↓kaQte i↑e↓nanie.
+nyaru/187.wav|2|yo↓ku a↑reda↓Njono ki↑moʧino bo↑rute↓eji.
+nyaru/243.wav|2|no↑re na↓nio ʃI↑teno i↓ma yo↑aanotakawaka↓atoya ko↑to↓dene ya↓i ma↑ʃi no↑rineisUki.
+nyaru/295.wav|2|ʧi↓suzuga ya↓Qta ko↑to↓ga do↓renake hi↑do↓i ko↑to↓kao ʃi↑Qte ka↓nojoni a↑yama↓Qte ʧI↑kazuka↓nai yo↓oni.
+nyaru/32.wav|2|a↑ase↓Qkaga ki↑ere.
+nyaru/260.wav|2|m i.
+nyaru/50.wav|2|ʧI↑ʧi bi↓i te↑Ntoʃi↓ayaga o↑retanawataʧi↓domade i↑ta yo↑koʃimada↓na.
+nyaru/31.wav|2|ka↑o da↓Qba wa↑ita.
+nyaru/399.wav|2|w a to↑me↓sude na↓kU ka↑teemo ke↓Qkoo sa↑soe↓fUkudesU ʃi↑ka↓ʃi do↑oʃIte↓mo sU↑ki↓no ko↑ga de↑ki↓ma.
+nyaru/355.wav|2|a↓nika ta↑noʃi↓i ko↑to↓wa ʃi↓aji na↑Qpa↓ritokani so↑NN ro↓ni ji↑kaNwa tsU↑kaQte.
+nyaru/62.wav|2|ya↓N da↓te ru↓karane ma↓anewa a↑wa za↑tsu mo↑to↓ka e↓mina sa↑gasetsu↓naitaa go↓owa mo↑to↓kaita.
+nyaru/55.wav|2|i ta↑N↓sakuni o↑ne↓gaigoto ka↓ku no↑wa ʧo↑saku↓kina ji↓ya z u g a ko↑refuriiN↓saitona ha↑kaijii↓desU.
+nyaru/366.wav|2|ta↑be↓ta.
+nyaru/166.wav|2|hi↑ragakaaateezuN↓tsUki a↓Qta ko↑ga de↑ki↓ta to↓kini ho↑mehatsude ʧI↑ʃiN na↓gu N↑dao ta↑yo↓ita.
+nyaru/300.wav|2|ti↑ʃimikatatasUte↓ʃi i ʃi↑ra↓ʃaNde o↑wa wa↑rewataritaiʃase↓nakani de↑kuremase↓Nka hya↑ku.
+nyaru/317.wav|2|ki↓ka yu↑zurute i↑ta ʃi↓i go↓orujana ʃa↓ne go↓meNno ze↑Ngomeno sa↑iozukIka↓nari.
+nyaru/264.wav|2|a↑jiga o↓mariwa pa↓tena.
+nyaru/342.wav|2|ha↑yaya↓kUku.
+nyaru/362.wav|2|ki↓ʃia ʃi↑mo↓sunayabuaga tsu↓kuQte o↓ro i↑ʃimahainanaka↓iʃayo.
+nyaru/78.wav|2|e i↑Qtasu.
+nyaru/147.wav|2|di↓suuNno kya↑akutaaura na↓i.
+nyaru/215.wav|2|Q pa↓ni ka↑waQte ku↑reto mu↑ʃirenaima↓ʃIta.
+nyaru/327.wav|2|k a wa↑sureru↓no u↓to mo↑oke↓no wa↑sureka ta↓Qte wa↓kana i↓ra ji↑kaN↓me.
+nyaru/53.wav|2|e↓ine yo↓kokuno so↑mata no↑saN.
+nyaru/199.wav|2|n i ke↑QkoNNde su↑ru mo↑no↓jina ʃi↓i.
+nyaru/270.wav|2|i jo↓kyooonoQtetotawa.
+nyaru/357.wav|2|na↓nie.
+nyaru/99.wav|2|re↓jie bo↓kuwa o↑Nna↓nokono te↓saio fu↑reta ko↑to↓ga na↓i do↑o↓hododesU bo↑ku↓roo.
+nyaru/137.wav|2|k a ro↓oga tsU↑kikiga↓dokakaga ru↑iʃi↓jiN.
+nyaru/140.wav|2|wa↑kuwa↓karekara mu↓jano a↑i↓naN m o te↓o to↑ri↓no ku↑raNdoa↓nanimo na↑ka↓Ndeena su↑sa↓ne.
+nyaru/70.wav|2|zu↓baawa a↓iisUto ha↑i↓ke ma↑ketoriha↓Ntai nu↑kasede na↓suno wa↑areN↓sega ji↓keNdewa.
+nyaru/43.wav|2|t e ta↓tanooseekio o↑Qte ki↓taga r a.
+nyaru/79.wav|2|o na↑N↓ʧano ʧa↑ari↓gatoo.
+nyaru/363.wav|2|ka↑i.
+nyaru/7.wav|2|na↑daaniisoni tsU↑ku↓Qta N↓da yo↑o↓gaku.
+nyaru/112.wav|2|hi↓yana no↑ga.
+nyaru/18.wav|2|n i hI↑ki ta↑i↓hoo ta↓neaikeNtoka ta↑nahanaʃi↓doni se↓nakaQta m a.
+nyaru/310.wav|2|na↑igawaga a↑reeru to↑ri↓koi a↓Q ʧa↓ae di↓ʧomo ha↓iQte re↓jiyaNNto ko↑mateno k o ma↓iyo m i.
+nyaru/142.wav|2|ji↑ue↓neN i↑nu↓mo sU↑ki↓Nwa a↑ru↓Nneyo.
+nyaru/19.wav|2|d e ʃi↑Qma↓i.
+nyaru/265.wav|2|de↓miinisutotowa u↑keseeno ke↓Nrio mi↑ge sa↑NNzunarimiʧIto ta↑yooseeo ʃi↑Qpuu.
+nyaru/13.wav|2|na↑nieeki↓ekIka.
+nyaru/284.wav|2|wa↑a r a s o ko↓e ha↑na↓su.
+nyaru/181.wav|2|m o ma↓dane ki↓te i↑mase↓N.
+nyaru/278.wav|2|u↑Nkora↓geN ko↑rogasa.
+nyaru/323.wav|2|mi↑se↓kara ka↓geNjanaino o↑toriga u↑tsUʃi↓meN.
+nyaru/332.wav|2|i↓ma ta↑e↓ru to↓kide sU↑kuna↓eru.
+nyaru/296.wav|2|ʃi↑ma↓ʃIta.
+nyaru/274.wav|2|ma↓riwano ʃi↑i ʧo↑ku↓kino.
+nyaru/42.wav|2|Q↓ta ma↑ʧi↓ne a.
+nyaru/376.wav|2|n i de↑aireQte na↑kanaka na↓ito o↑mo↓u.
+nyaru/196.wav|2|ma↓demo a ʧo↓Qtodemo a.
+nyaru/127.wav|2|tsu↑raʃIte.
+nyaru/283.wav|2|ji↑buNga ʃi↑yosegaraiNja.
+nyaru/194.wav|2|ki↑moʧina o↑okIsaNni↓saga a↑repata↓aNno me↓e mi↓kaga.
+nyaru/178.wav|2|na↓sa i↓eeta sa↑kigoro ha↑kosakI ka↑re↓jina ki↓tsuge ʃI↑ta↓ra a↑su↓ga.
+nyaru/48.wav|2|wa↑tade↓sU de↓NdeNka de↑Nwa ʃI↑ta↓kara.
+nyaru/66.wav|2|e↓N.
+nyaru/47.wav|2|da↓a.
+nyaru/285.wav|2|a↑reji↓buN na↑ʃiyaasegaraiima↓sU.
+nyaru/287.wav|2|n i.
+nyaru/220.wav|2|i↑niʃie↓tekiniwa ʧo↑ofUtsu↓koN ka↑ramake ko↑N↓oʧina wa↑ka↓i se↓eʃiNo ge↑tanoni ki↑ri↓koN.
+nyaru/121.wav|2|fu↑rareru↓noo ma↑ta ke↑ekeNnokana.
+nyaru/319.wav|2|tsu↑gaegoe↓eno e↑ko↓ewa ji↑sUkikaga to↓gainasai go↓eN na↑sa↓i.
+nyaru/381.wav|2|ʧo↑okeeyareʧaN wa↑taʃimo ni↓seN.
+nyaru/252.wav|2|n o mi↑ʧijoomo to↑bu↓na ma↑maʃi me↑ʃi↓raga yo↓mina.
+nyaru/188.wav|2|.........
+nyaru/186.wav|2|na↑NtaNʃii↓neyo.
+nyaru/122.wav|2|g a te↑natewadoʃItani we↓ʧi sa↑a ma↑arete↓kiane re↑e↓ai ʃi↑nai↓kara ra↓Nkayo.
+nyaru/119.wav|2|a↑ataNheNko↓oda ʃi↑nai↓kara ka↑reʃaga de↑ki↓na e↓naN.
+nyaru/403.wav|2|na↓i ka↑nooseewa a↓ru.
+nyaru/249.wav|2|ya↑ʃi↓ina ʃi↓N da↑ma↓Qte ku↑re.
+nyaru/192.wav|2|tsu↓komadeN.
+nyaru/160.wav|2|go↓toe ʧ i ke↓Q te↑ra↓ne.
+nyaru/246.wav|2|.........
+nyaru/259.wav|2|te↓N ni↑jo↓na ta↑mago↓goeo fu↑mi↓Qta wa↑taʃI te↑N↓seN mi↓o tsU↑kaQta te↓N fU↑tari.
+nyaru/110.wav|2|tsu↑maseNjooʃa su↑re↓ba i↓naaQta do↓ga te↑gaogaQte o↓i no↑wa s u.
+nyaru/204.wav|2|ʃo↑okai hI↑to↓rino hI↑to↓bitotoo ke↑QkoN ʃI↑te↓saga ma↓ki i↑keta↓ra i↓ʧiwa i↓itanasa.
+nyaru/405.wav|2|e↑kakerare↓ru ki↑ga su↑ru N↓dane na↓su o↑Nna↓na ko↑ʃirore↓naite i↑nata ka↑Qteni ha↑naʃIkakerareruita↓ina.
+nyaru/81.wav|2|ta↑re↓dayo pe↑eta↓aruta yo↑N↓teNNmao re↓wa.
+nyaru/174.wav|2|i↑ta.
+nyaru/349.wav|2|i↑sogaʃi↓QkUsurukamo ka↑Ngae↓ra hi↑mamo na↓ino ra↓i.
+nyaru/386.wav|2|ts u o↓bini ka↑NʃItewaa ma↑areNtaisurunaNkade i i↑ya↓garasegaQta ni↑ru ma↓rega ko↓ru ya↓gaga.
+nyaru/353.wav|2|ka↑ra na↓i yo↓oni ʃI↑ta ho↓oga a↓tsurai ji↑kaNwane.
+nyaru/184.wav|2|n a re↑o↓hono to↑ʃi↓korewa re↓a ke↑ekeNsoodaNsoodaNma↓ʃuugaQta.
+nyaru/161.wav|2|sU↑ki↓ya ma↑i↓teo ni↑gigireno na↓ku mo↑Ndainakima↓sene.
+nyaru/340.wav|2|m a.
+nyaru/226.wav|2|ts u so↑koiewa ko↑o↓naito i↑ʧi↓daini i↑enaike↓no.
+nyaru/280.wav|2|na↓ku na↑ni↓waa.
+nyaru/347.wav|2|ma↑gorikigo ʃI↑ta↓surani i↓N.
+nyaru/103.wav|2|tsu↑gi↓gaojanayo bu↓N do↓waa ti↑de ra↑ku↓zu ka↑Njigawajoonake te↑ra↓Qtamo ʧi↑a↓ʧiaekoʧi.
+nyaru/145.wav|2|.........
+nyaru/67.wav|2|re↑ga↓moto yo↑ri↓yukino ka↑mi↓no ha↓i ʧI↑ka↓desU.
+nyaru/222.wav|2|N↑ie↓eNni o↑mo↓u ki↓mo a↑reka↓mo ʃi↑renai↓karasa.
+nyaru/401.wav|2|ko↑rewane a↑nomado↓rigano ki↑mino ka↑oo mi↑te↓naikara na↓Ntomo i↑enai ke↑ta.
+nyaru/322.wav|2|a mu↑ʧiro↓N ji↑buNnate ʃi↑ma↓Qta hi↑ro↓i ko↑to i ʧa↑Nto na↓NkaQte a↑yamaQta↓ri.
+nyaru/86.wav|2|a↑reaQtaori↓are a↓ru na↑kako↓kUhakUto k a su↑ru hi↑gano mu↑ne.
+nyaru/167.wav|2|ka↑Qta ha↓riyuaʧI ʧi↑ija↓ku wa↑tate↓NʃIto ʧa↓kuga i↑naku na↑sa↓i.
+nyaru/14.wav|2|no↓de ko↑yariaiʃimaʃItao i↓ta no↓mu ʧo↓to mu↓ʃiro i ke↑ekeNdaQ pa↑ri↓toka.
+nyaru/239.wav|2|ko↑mi se↑N↓ma o↑reya ji↑makIʃiN↓ga mi↓ru no↑ga da↓isUki mi↓ta ko↑to↓ga i↑dakenu.
+nyaru/162.wav|2|i↑iNʃa↓kai i↑de ʃa↑Nka.
+nyaru/275.wav|2|ri↑na↓jiN so↑reN↓zore.
+nyaru/233.wav|2|me↓mo ji↓Nseena ka↓naN.
+nyaru/37.wav|2|i↓ma s u.
+nyaru/21.wav|2|e↑ki↓kieru e.
+nyaru/245.wav|2|n o jo↑oki↓teN.
+nyaru/333.wav|2|ko↑re na↓ma i↑dera↓ewa ko↓raidesUka na↓izo i↑mawa.
+nyaru/391.wav|2|d e ha↑NseeʃIte sa↑kao tsu↓koQtemota fu↑ro↓ʃI ki↓tekUsai.
+nyaru/277.wav|2|ma↓ga do↓rega i↓ito ma↑nabe↓ai i↓sui a↓kirawa na↓ina.
+nyaru/367.wav|2|s a so↑rage a↑tsUkuQtee fu↑ru↓e je↓eQ ta↑Qta.
+nyaru/87.wav|2|ki↓o ke↑gare↓raga k i a↑ʃi na↓i o↑oijami↓mee ma↓areeno ki↑ite i↑ta↓ga.
+nyaru/352.wav|2|de↑Qte r u ji↑kaNga i↑ʧibaNka↓bini so↓mukeretano na↓kana me↑eka↓kaN ʃi↑ro↓naN ji↓ka.
+nyaru/324.wav|2|ka↑ra yu↑ru↓ʃIte ku↑re i↓to o↑mo↓u.
+nyaru/30.wav|2|tsu↑riwa i↓ma ki↑zu↓ita.
+nyaru/164.wav|2|so↑no to↑ki↓nina ʧu↑udaiodoʃIta↓ri o↑riree a↑ji te↑ta↓i ni↑giru N↓daya me↓N to↓ka ra↓Nkao i↑u.
+nyaru/94.wav|2|za↑N↓mina za↑NʃiNna a↓i o↑keni ʃi↑yoo.
+nyaru/152.wav|2|ma↑ma↓demone a↑nowa ko↓miʧina ha↑iri ne↑kota↓gainu wa↑ru↓i ho↑ne↓demo na↓ikarane a↑noha↓Ntai ta↑noʃi↓Nde.
+nyaru/183.wav|2|to↑bu sU↑ki ya↑rete kU↑ʃa↓ri i↓NjaQtaga.
+nyaru/2.wav|2|k i gu↓ragura ma↑imi ra↓Nde sU↑kero↓wa.
+nyaru/27.wav|2|na↑wa↓tateN.
+nyaru/150.wav|2|ze↑Ntotsu ʃI↑tena sa↑i↓kyoono kya↑ani ʃI↑tene ju↑u↓neN a↑soNda ho↓oga i↑ru.
+nyaru/83.wav|2|mi↑se.
+nyaru/354.wav|2|ra↓sereto o↑mowa↓nade.
+nyaru/397.wav|2|hi↑roʃiji↓ʃiNwa a↑reNta↓ne wa↑taʃI ʃi↑o yo↑mi↓ʃItsuni na↑N↓kyona r i ki↑N↓eNNte ko↑to↓ka.
+nyaru/5.wav|2|a↑sa↓ʃao e↓ta go↓o da↑Nʃaa↓deN ʧa↑wa sa↑Npao ʃI↑tsugoe↓iN.
+nyaru/385.wav|2|t a bu↑Nji↓doode bu↓rokU sa↑reta ha↑naʃi↓to o↑moima↓sU.
+nyaru/10.wav|2|N↑wae ku↑darita.
+nyaru/224.wav|2|i↓ma.
+nyaru/131.wav|2|o↑mo↓Qte sU↑ki↓aga a↓Qte o↑mo↓Qta hI↑to↓yoogae de↑ki↓ta to↑kiri↓iNja a↓i.
+nyaru/247.wav|2|.........
+nyaru/17.wav|2|ho↑ʃi↓ine na↓i yo↓owa nyu↑utatetara↓nio ke↓sarano ta↓e ku↑dari su↓ita ma↑ta ma↓tsu.
+nyaru/46.wav|2|a↑eyami↓ʧibu ji↑Qto u↑tsua↓sarataga a↓Qte.
+nyaru/193.wav|2|ki↓deara ya↑su sU↑ki↓nakedo so↓o ho↑te↓Qteyo.
+nyaru/98.wav|2|ne↓ene.
+nyaru/104.wav|2|ʃi↓aN yo↓rani ba↓kani tsu↑gewa i↑te ko↓tawa.
+nyaru/404.wav|2|to↑yanamugo↓toʧihoʃizukIʧino i↑ʧi ki↑menaQta↓ga ma↓rikara.
+nyaru/380.wav|2|ki↓riʧiN a↓rita a.
+nyaru/286.wav|2|na↑ge ra↓N a↑Nmari a↓Nka ko↓ogamo ko↑meNto↓ʃina u↑se↓so.
+nyaru/134.wav|2|na↓no ne↑e↓aa ʧu↓u ha↓NN ta↑isUkino ne↓koga de↑ki↓ta go↓uta su↑ne↓wa.
+nyaru/382.wav|2|n a na↓kani bu↑roQkUsa↓rede na↑kima↓ʃItaQte go↑meN↓sasaiwa.
+nyaru/227.wav|2|yo↓mo m a.
+nyaru/231.wav|2|ma↓to ko↑koni ke↑QkoNsa↓Qʃi mo↓o fu↓meNdewanaikedo bu↑N↓jutsu ya↓gitakaQta ko↑to↓toka a↓QtanoninaaQtoka.
+nyaru/157.wav|2|to↑ʃi↓heyani ya↓Qtasu.
+nyaru/107.wav|2|sU↑ka ma↓nanatowadene.
+nyaru/321.wav|2|ko↑jiiNgyoohawa hI↑kidaʃIta ka↓Qtotieno a↓a ʧo↓QtoʃI↑temo ʃi↓Nikedo.
+nyaru/258.wav|2|ni↑mupemi ni↓teNnu.
+nyaru/205.wav|2|ta↓kamo ʧI↑ku da↑ijini ke↑QkoN sa↓Qte o↓itemo i↓i N↓janaide ʃa↓youka.
+nyaru/314.wav|2|ma↑mizuinameme↓gakUʧaN e↓de ʧo↓Nni u↑reteNNreja↓rona↓gaQte ʃI↑ta.
+nyaru/175.wav|2|ju↑u↓ʧa tsu↑gi.
+nyaru/348.wav|2|ho↓Nya ka↑tsuzuku↓ri ʃI↑ta i↓iii.
+nyaru/369.wav|2|sa↑ta.
+nyaru/190.wav|2|mo↓ʧi re↓ʃiNno sa↑ita↓ina.
+nyaru/219.wav|2|na↑asaNdaiwa ni↑juusagariha↓Nno ka↑na↓Qtetana sa↓a ʧo↓Qto ha↑naʃi↓o ka↑Qte ku↓riyonesa o↑Nna↓no.
+nyaru/201.wav|2|m a kyu↑u↓niNni yo↑rukamo ʃI↑te na↓ki no↓saaa ma↑iʃanita ta↑tekoN↓joodi ko↓N ka↑wa.
+nyaru/29.wav|2|ke↓etoka ka↓eru yo↓kaQtaga.
+nyaru/326.wav|2|ʃi↑N↓ruito ta↓tsuiNda Q↑pana.
+nyaru/293.wav|2|ta↑meka↓nojoni a↑puro↓uʧi ʃI↑ta↓ri ko↑jiNjo↓ohooo hI↑kidaʃIta↓kaQta.
+nyaru/114.wav|2|u↑ʧiraN↓ʧI su↑ka ni↑Qpo↓Ndayo te↑eja↓faa.
+nyaru/297.wav|2|de↓mo i↓mademo ki↓doki mu↑ne↓ga i↑ta↓idesUwa kI↑souni na↑rima↓sU.
+nyaru/3.wav|2|a↑no↓osusaN de↓rosu.
+nyaru/305.wav|2|d e ko↑magoreni ta↑iko i↑Qtenu no↑boreN.
+nyaru/371.wav|2|i↑naareʃa wa↓reda ki���jina↓iʧi no↑boQte nyu↑uʃiNNji↓teN ra↓iNna no↑ka p a.
+nyaru/69.wav|2|ho↑ri ba↓Nwa ryo↓o ʃi↑reta N↑na po↓i su↓mu.
+nyaru/172.wav|2|tsu↑morina N↓desUkenoo.
+nyaru/92.wav|2|mi↑Nna↓da.
+nyaru/200.wav|2|to↑a.
+nyaru/292.wav|2|de↓mo ka↓nojowaga ʧI↑koi↓joo i↑ranairaʃi↓kUte so↑rede↓mo zu↑Qto ji↑buNno ko↓ega mi↓ru.
+nyaru/169.wav|2|na↓gewa ra↓Nkao wa↑kaʧi↓de ha↑tsu↓ya te↑noowa ka↑N↓naNka na↓i to↑me↓ma.
+nyaru/25.wav|2|w a i↑Q↓kai ʃo↑okoma↓mane.
+nyaru/289.wav|2|e↓ta.
+nyaru/364.wav|2|mi↑se.
+nyaru/281.wav|2|re↑nai ho↓ga i↓iyone fU↑ku↓ʃi.
+nyaru/244.wav|2|da↓u.
+nyaru/80.wav|2|ma↑wariga ho↑ʃi↓i da↓da y a.
+nyaru/141.wav|2|ki↓ga go↓runo re↑osaabaroraatomusuʧIkaʃo gi↑su↓rebasaa.
+nyaru/23.wav|2|a↓ekie ta↑nan i su↑redamata de↑ki↓rukana.
+nyaru/115.wav|2|re↑emia↓ru ji↑i pi↑i↓iN ma↓areeuurujiapiide sU↑ʧaa↓piN.
+nyaru/308.wav|2|mo↓reni mi↑Nnaemorewa gya↑kurigyakuiNN↓meene a↓ru ji↑i mo↓o re↓ni ho↓rere ma↓e.
+nyaru/165.wav|2|d o de↓sa i↓karanakU ʃi↑reQ kU↑tsu na↓ge ha↓iitoo.
+nyaru/368.wav|2|.........
+nyaru/212.wav|2|ru↓wane.
+nyaru/88.wav|2|ʃo↓oga ʃI↑kaiQte ji↑keta↓jiN u↑so↓ijaNki no↑boQte y a yo↓mo.
+nyaru/279.wav|2|do↓rega se↑e↓gi te↓Qtara da↓rekaga te↑kini na↓rukarasa.
+nyaru/254.wav|2|.........
+nyaru/51.wav|2|to↑ko↓ri ʧa↑Nto ko↓ra.
+nyaru/144.wav|2|wa↓kariwanaa.
+nyaru/331.wav|2|ha↓desU.
+nyaru/339.wav|2|a↑nowa.
+nyaru/168.wav|2|te↑Qʧire↓Qtotsu na↓ine.
+nyaru/151.wav|2|ya↓nayana.
+nyaru/35.wav|2|wa↑tatana↓oNno ha↑ikeea↓koretakae i↑ta↓na.
+nyaru/217.wav|2|t e mo↑sareeageno ʃ aka↑mi↓guN re↑Nmetoka↓Nni na↓Qtete e.
+nyaru/251.wav|2|se↑N↓neNwa.
+nyaru/267.wav|2|i↑Ngaanohaieaaʧisusuʧuuni se↑mii↓ʃItsu de↓ru ko↑to↓boohyaN ʃI↑tani.
+nyaru/335.wav|2|u↓N.
+nyaru/225.wav|2|a↑no ki↑mini tsU↑kaQte ku↑rete i↑ru no↑kana.
+nyaru/393.wav|2|ni↓ji ni↑sai↓naNdesUke no↑wa o↑Nna↓nokono te↓o ʧ i na↓gu ko↑to↓sura de↑kimase↓NdeʃIta ba↑akori↓yukiina.
+nyaru/341.wav|2|do↓oʃIteno wa↑sureta↓i na↓Qtara be↑tsuno i↓saNo o↑ʃIsuQte i no↑ga i↑ʧi↓baNnaNka pe↑Qto↓ri.
+nyaru/242.wav|2|na↑na↓kai ʃ i.
+nyaru/28.wav|2|na↓i hI↑toteN↓roNka.
+nyaru/389.wav|2|t e ku↓dosae.
+nyaru/0.wav|2|ʧi↑gakUheesaNki↓sa ke↑suga.
+nyaru/343.wav|2|ba↓suni ko↓ori he↓Njaga o↑konaikazo.
+nyaru/398.wav|2|g a ge↑N↓gakU ko↑to bo↑o do↓katokana.
+nyaru/146.wav|2|i↓ika ha↓ra mi↑ni ko↓QkI ʃi↑ro↓i no↑ji ge↑Nsusumumie.
+nyaru/298.wav|2|d o a↑ʃIta↓wa ka↓nojoo wa↑surerarema↓sUka o↑ʃiete ku↑dasa↓i.
+nyaru/108.wav|2|wa↑zameNfa↓atani su↑ru.
+nyaru/138.wav|2|o↑kane o↓kooni tsU↑kaQta↓awa gi↓janai ba↑rora↓N to↓N N↓sUki ka↑Q pa↓wa gi↓janai.
+nyaru/210.wav|2|ko↓nakino ke↑QkoNdekioto ʃI↑ta.
+nyaru/124.wav|2|ka↑Qta↓aya a↑yama↓ʃimasetewa i↑itsuuwa.
+nyaru/1.wav|2|u↑ra↓ni ge↓No sa↑iʃo ka↑rage↓nageda ra↓Nde sU↑ki↓no a.
+nyaru/52.wav|2|ʧa↓N N↑to go↑rahora↓nana ta↑deyo go↓ra s u.
+nyaru/195.wav|2|ra↓me z a ka↑ʃiiyone.
+nyaru/128.wav|2|a↑rai.
+nyaru/209.wav|2|ʃa↑kaiji↓Nkatowa mi↓raidakedowa.
+nyaru/373.wav|2|te↑te.
+nyaru/365.wav|2|d e sU↑kuQta.
+nyaru/71.wav|2|tsu↓mino de↓Nki no↑marenu ha↓tsuwaʃIka ji↑koiakyoo to↓rinaQtede.
+nyaru/372.wav|2|ku↑royaQta↓kU fu↓na i↑ki↓donee ya↓Qta ko↓naki no↑o ʃi↑aranetaraasoNwaza tsu↑ru↓wa mo↓re.
+nyaru/303.wav|2|ha↑ʧi ʃ i yo↓rurigatoo a.
+nyaru/266.wav|2|se↓reʃItano ko↑to↓de.
+nyaru/95.wav|2|mi↑gini.
+nyaru/61.wav|2|ki↑ni sU↑tekara wa↑rearaitokunemuri↓dooete ʃi↑ra e↓te ʧa↓N tsu↑etsUteka↓su.
+nyaru/346.wav|2|w a ka↑mikuya go↑neNnetsu↓mazo re↓eraQtara do↓o wa↑sureru N↓daroona.
+nyaru/253.wav|2|da↑tesaku↓biN N↓faata.
+nyaru/301.wav|2|yo↑ika↓radana ha↑i↓seeo a↓rega te↓i to↑ri↓togeo ku↑re u↑ʃIta ya↑tsu↓jaamezuraʃi.
+nyaru/213.wav|2|ko↑magawase↓eto ka↑na↓Qtara ʃI↑ka↓ʃi na↓i ke↑QkoN ʃi↑Qte to↓moo ʃI↑te da↑iga↓kUsee.
+nyaru/73.wav|2|tsU↑kuekurekatawawadeN↓bi ʃi↑ma↓sU ka↑i k u jo↑oomenamajooya ha↑i↓ʃiNde be↑Nkyoose↓seki j o tsU↑ku↓rodeʃIte.
+nyaru/315.wav|2|mo↑esesete na↑Npi↓suno a↑ya↓ri ji↓maNre nu↑enu.
+nyaru/290.wav|2|ma↓re.
+nyaru/56.wav|2|ra↑furiiso↓zaidesU ko↑mo↓deN se↓ita no↑wa.
+nyaru/75.wav|2|fu↓ʃItsu fu↑Qta↓iyo.
+nyaru/318.wav|2|w a ʃI↑ta tsU↑ʧi↓deN i↑nu↓saN sU↑ka↓Qta.
+nyaru/191.wav|2|ha↑ʧi↓kana yu↓o ka↑nojo↓yamano ho↓ga ʃI↑kiʃI ki↓te tsu↑mamiseekuN ma↓da.
+nyaru/41.wav|2|ge↓rutone ko↑reoo ka↑Nga↓e wa↑ta.
+nyaru/189.wav|2|ʃI↑ʧooʧinorazume↓etaano ta↑ka↓saga a↑wa↓na ka↑Qta to↓kino.
+nyaru/240.wav|2|N↓wa do↑buNwa sa↓sune.
+nyaru/334.wav|2|ha↑ra↓i ke↓NʧIka na↓i su↑ne.
+nyaru/379.wav|2|ha↑itsugieeekoʧirani↓iʧaNwa yo↓ru na↑mi↓haʃi.
+nyaru/344.wav|2|o↑moima↓sU.
+nyaru/44.wav|2|ʃo↓ki na↓ʃIte go↑uta↓eNkara ko↑bore.
+nyaru/262.wav|2|d e i↓miga wa↑ka i↑ʧiure↓teN.
+nyaru/234.wav|2|fu↓Ntaani.
+nyaru/395.wav|2|mo↑ʧi↓roNga ya↑Qtsu↓koo ʃI↑ta ko↑to↓mo ha↑ʧi↓koo ʃI↑ta ko↑to↓mo na↓kUte e.
+nyaru/387.wav|2|do↑mo↓Qta ʧo↑o↓deN ta↑oʃItani na↑Nkamo na↑N↓kaao ʃI↑te ki↓ta hI↑toa↓kUʃi te↓ruto o↑moima↓i.
+nyaru/156.wav|2|ki↑reku↓na re↓taa kyo↓o i↓makara ki↓N ta↑me↓su yo↓one ma↑warini ke↓tsU su↑tete ki↓te r u ʃI↑te i↑rudeni ya↓Qta.
+nyaru/59.wav|2|sa↓itonaa.
+nyaru/216.wav|2|ne↓N ʧo↓Qto.
+nyaru/263.wav|2|go↑ʃii ko↓keni pa↓ne na↓ze i↑Q↓see ra↓N ʧ i ya↓ra ka↓Q ʧa↑radake.
+nyaru/351.wav|2|ya↑ba na↓nimo ʃi↑nai ji↑kaNtoka to↓kuni so↑no ne↓re ma↓eno tsU↑tomuNde a↑negena he↑ra.
+nyaru/229.wav|2|ʧe↓to ʧo↓Qto wa↑raN↓iNo↑Q ʧ i ji↓iyoQteno su↑nanoni ha↓iQta h o e↑iiNʃa↓kedesU yo↓uka.
+nyaru/40.wav|2|n o.
+nyaru/271.wav|2|ni↑ke↓Ndeno de↓baNde mi↑o no↑Qte i.
+nyaru/102.wav|2|gu↑muka↓Njie u↑e na↓kaQta u↑ma↓o N su↑re↓wa i↑te ko↑ta↓ku.
+nyaru/9.wav|2|w a i↑ʧiniʧi↓kai to↓nekonedesUne.
+nyaru/312.wav|2|ma↓areedewa na↓i ro↑kUkoma↓oreno ko↑to↓mo ʃI↑kasete.
+nyaru/235.wav|2|wa↓nano ka↑Qte ʧa↑Nto ka↑Nga↓etekara ke↑QkoN ʃI↑ta ho↓oga i↓ita o↑maemase ne↑ko↓rawa.
+nyaru/268.wav|2|pu↓roQta i↑tsUkuʃibiyani ki↑re↓ru ke↓Nde s u to↓i ʃI↑ta i↑su a↑raware↓tede sU↑ko↓ʃi bu↑Nji↓ʧI sa↓ite.
+nyaru/132.wav|2|ka↓i na↓i to↓oo ya↑wara↓i ka↓no ʃi↑renai mi↑te↓Nga.
+nyaru/6.wav|2|ne↑gai↓ʃiNde a↑gemaʧo↓oha.
+nyaru/325.wav|2|ma↑tagano.
+nyaru/171.wav|2|e↑emaareNseNseeboku↓kano ka↑re↓jiwa to↑i su↓gu ni↑Qpo↓Nna ri↓gakU se↑ekatsuga o↑waQte ʧ in i mo↑do↓ru.
+nyaru/117.wav|2|ka↓nojosura i↑naite kU↑ta↓wa.
+nyaru/72.wav|2|ne↓Nrai sa↓Ngatsuni mu↑kete ka↑N↓jide ta↑mote↓ʃi u↑Ngaga ma↑oyowaiseeseena ka↑No↓Nna.
+nyaru/272.wav|2|na↓Nda.
+nyaru/338.wav|2|n o su↑sa ke↑QʃIte ku↑reru↓ka.
+nyaru/356.wav|2|i↑kima↓ʃi ya↓a.
+nyaru/149.wav|2|y o to↓tsu ʃI↑tene.
+nyaru/218.wav|2|a↑temo na↑N↓neNmo ku↓etete o↑Nna↓na ko↑to↓kara to↑ʃi↓mo ni↑ji a↓ru.
+nyaru/307.wav|2|i↑Qta.
+nyaru/16.wav|2|pa↓naabatayao mi↓ito go↓zama ʃI↑te i no↑ga wa↑taQ↓ku na↑Ni↓ne.
+nyaru/241.wav|2|ni↑gyoo↓kiNga a↓rukarato ko↑to↓wa.
+nyaru/374.wav|2|ni↓hatsU ku↑raQpanaaakaha↓aN pa↓ani ya↓demo s o ne↓te re↓e i↑kaketagena sU↑ki↓na hI↑toga de↑ki.
+nyaru/170.wav|2|su↑gi↓e.
+nyaru/159.wav|2|ni↑giQta ko↑tsu na↓i no↑ka.
+nyaru/378.wav|2|.........
+nyaru/84.wav|2|a↓kawaga o↑ʧi↓rukana ba↑reNta↓iNgewa hi↑jooni ka↑Njoo↓tekine u↑ma.
+nyaru/113.wav|2|e↑so↓seeku to↑Q ʧ i na↑u↓aQtene.
+nyaru/236.wav|2|te↑tenonegai su↑ru↓na ʧi↓ni ʃi↑naiga na↓maneN.
+nyaru/36.wav|2|Q↓pa hi↑rakU ʧ i te↓ka.
+nyaru/261.wav|2|se↑miimiisutoQte ko↑to↓ma.
+nyaru/154.wav|2|ka↑wa ha↑ʃi ʃi↑ni da↑i ko↑ora↓aaQtemo ʃI↑ta↓ka de↑ki↓tara.
+nyaru/105.wav|2|ʧi↓no ʧI↑te ka↑Njiwae ta↓kide ka↑N↓jiga a↓ito a↑te.
+nyaru/223.wav|2|to↑ri to↑narikai↓sano o↑Nnarokono ji↓Nseeo.
+nyaru/33.wav|2|o↑otsUkiga yu↑Njaratairaji na↓i u↓eni.
+nyaru/68.wav|2|so↑no ba↓ano se↓eki a↑taerema↓sUka.
+nyaru/129.wav|2|ts u ra↑a↓u ni↓saN.
+nyaru/304.wav|2|y a ku↓u.
+nyaru/60.wav|2|so↓rananode ra↑ihebio↓QtodesU no↑itoguʧi↓sakU ka↑naga↓ʃira mo↑ri m e tsU↑kio ʃI↑te↓rukara o↓otoki.
+nyaru/4.wav|2|i↑makoraʃa↓yooeto ʧI↑kiiNzuʧaNtsumamizeegotoʃayayokuyamaha↓No.
+nyaru/276.wav|2|i↑roNna ka↑Nga↓ega a↑rekaya.
+nyaru/97.wav|2|me↓nu ma↑marokoʧira↓eN.
+nyaru/130.wav|2|a↑yate↓e s o ʃi↓zuto ʃi↑nirutokamo yo↓ku wa↓kano i↑sUkaji↓buNa ko↓N to↓oni tsU↑kiaita↓ito.
+nyaru/337.wav|2|ji↑kaNtone a↑taraʃi↓i ko↓ega.
+nyaru/392.wav|2|ha↓i d e ko↑ʧi a↑rai ʃI↑harinobose a↑raimasU e↑eQta↓ji y a ʧa↓N wa↑taʃI ʃ i.
+nyaru/64.wav|2|to↑o↓sakujanai o↑ko↓re ʧa↑Nto ki↓ga to↑maretema↓sU.
+nyaru/38.wav|2|i↓mata u↓yodesU.
+nyaru/155.wav|2|a↑no ze↑Qtaini ko↓okai e↓nai yo↓oni a↑taQ↓kuʃItawaito o↑mowa↓su.
+nyaru/90.wav|2|Q↓ta da↑i↓takega ʧi↓ʧiNja ha↑ra↓reni ha↑neni ka↑sune↓e.
+nyaru/133.wav|2|na↓da su↓re ra↓ka be↑tsuni ya↑wara↓ito o↑mowa↓naku i↓ito o↑mo↓i.
+nyaru/256.wav|2|e↓reka.
+nyaru/306.wav|2|ma↓rega ʧi↑ko↓ede ma↓ite i↑toQke.
+nyaru/143.wav|2|Q↓ʧiga to↓kUka.
+nyaru/320.wav|2|o↑otoʃIkotaeutoori↓ji ga↓buiwa ʧi↑rakana su↑e e↓Qtaa.
+nyaru/136.wav|2|y a mu↑ne↓deʃIte tsU↑kiahoga tsu↓iyo ʧi↑Qw a.
+nyaru/96.wav|2|i↑ni a↑ma ji↓wadesU.
+nyaru/402.wav|2|u mo↓ʃIka ʃI↑ta↓ra o↑re i me↑o↓toko ze↑Nteoraga hi↑ro↓reNni tsU↑ki↓ʧaQte mo↓te.
+nyaru/388.wav|2|ma↑NnakanaN↓nimori yu↑ruʃi↓ni bu↑ro↓QkU su↑ru ko↑to↓wanaikara mo↓ʧi ji↑buNno ku↑dari↓nooanon i ka↑e↓Qte m i.
+nyaru/329.wav|2|su↑eta↓mo ji↑kaNka ʃi↑N↓ra N ʃi↑i i↓maya a↑no ki↑zuo i↑ya↓ʃIte ku↑reruto o↑mo↓uno.
+nyaru/384.wav|2|ma↑ta bu↑eneji↓imaadotsuitaka na↓N ka↑re↓e.
+nyaru/345.wav|2|.........
+nyaru/126.wav|2|na↓ga i↑ru↓jaana ka↓tsuno a↓ya mi↓tsuga ha↓iru ho↑oda yo↑i mi↓taina ʃi↑mu↓koono hI↑tode ma↑waremeguze↓somo.
+nyaru/89.wav|2|ma↓niaa i↓makara ra↓N su↑ru kyu↑ukyuuo o↑kure↓ju se↑ki↓kio to↑re o↑samena↓iʧi.
+nyaru/269.wav|2|u↓mi u↓tsu.
+nyaru/109.wav|2|u.
+nyaru/45.wav|2|ri↑toʃijanai.
+nyaru/311.wav|2|.........
+nyaru/26.wav|2|de↑Nbutsude kI↑tawa↓taridesU.
+nyaru/358.wav|2|bu↑raQkuri↓sUtoni i↓rene da↑re↓tara do↓o ʃi↑ma↓sUka a↓o ko↑rene e↑ra↓Qkure↓ee o↑ʃi.
+nyaru/20.wav|2|e↓kieta ma↑rumaQ↓tatokanoba yo↓kaQtakana.
+nyaru/179.wav|2|n i ke↑QkoN ʃI↑tehoʃiito yo↑rema↓ʃIta.
+nyaru/54.wav|2|n a.
+nyaru/163.wav|2|na↓kadewa s a te↓ebigiQta ko↑tora i↑nai N↓da ki↑no↓saa ha↑ji↓mete mi↑serene↓tokI ka↑ma↓u ko↑no sa↑kI ku↓ruto ʃ i.
+nyaru/135.wav|2|i↑mare↓naku be↑tsuni.
+nyaru/207.wav|2|a↓Qtete na↑gare ga↑kUsee↓daQga o↑waQte mo↑no↓Qte ku↑renakara a↑kokooʃeka i↓ma.
+nyaru/85.wav|2|o↓bij a.
+nyaru/74.wav|2|da↓yameta ho↓ga i↓i i↑ya.
+nyaru/230.wav|2|i↑Qta ʧI↑ʧi↓iNNjanaideʃooka ko↑koe na↑garesakI ko↑i↓ʃini i↑mazumino ma↑rena.
+nyaru/39.wav|2|r a p a se↓ka ku↑raʃI sa↑rao ta↑ya↓suneka i↑so↓ita ta↓ga nyu↑u↓ki na↓su.
+nyaru/273.wav|2|a↑kabuno na↑na↓hari i na↑naʧuu so↑rezureno ka↑ra↓ni ma↑warina ko↑Qpe↓suni de↑merema↓ite o↑moima↓sU hi↓roʃItata.
+nyaru/350.wav|2|n o i↑toʃi k u ʃI↑te m a wa↑surerukamo ʧi↑rema↓i.
+nyaru/211.wav|2|ko↑okoo↓ka da↑igakU ka↓kU ʃa↑kai↓jiNnaegato u↓Nda ke↓roo i↓ma da↑sunu to↑ʃIkyooya.
+nyaru/375.wav|2|te↑i i↑kodayone su↑nani mi↓nokoQte mo↑so ni↑ʧi↓kina re↑re ʃ i.
+nyaru/182.wav|2|so↑Nna bo↓kuga do↓o ya↓Qte ka↓nojoto mu↑kaiae↓ewa i↓idesUka.
+nyaru/77.wav|2|N.
+nyaru/180.wav|2|ka↓ʃini to↓Qtewa ke↑QkoNni ta↑isu↓ru ko↑ko↓rono ju↑N↓ridemo se↑ekatsuni ta↑isu↓ru ju↓Nbide.
+nyaru/12.wav|2|ma↑ʃimaroga mu↑giikano re↑Naita↓ikeNdaQta a↓nie se↑N↓aini ta↓N ji↑kaN su↑rumi↓na.
+nyaru/176.wav|2|a↑Qʧi↓node.
+nyaru/228.wav|2|da↓kara to↑ite r a ka↓sa do↑aijoode ke↑QkoN su↑rere o↑kawaisoo↓aN ga↑ra ya↑bu↓e.
+nyaru/49.wav|2|ta↑nawa ta↑re.
+nyaru/91.wav|2|ha↓yani ya↑maito.
+nyaru/390.wav|2|i.
+nyaru/316.wav|2|de↓ne.
+nyaru/24.wav|2|ʧa↑Qto ma↓QteN.
+nyaru/139.wav|2|gya↑raNte↓sa ta↑kuiwa re↓to ʃI↑te ka↓Nzuni o↑kaneo kI↑tsu↓kaQta do↓ni ʃI↑te sa↓a.
+nyaru/328.wav|2|ke↓tsU tsu↑reta ma↑jide a↑no i↑roNna hI↑toga i↑ʃokuna to↑korode ya↑Qte↓rutemo ke↓do ma↓ʧi ji↑kaNga da↑i↓ketsu.
+nyaru/238.wav|2|na↑Njo ho↑dokose↓okU ko↑ko↓ro u↓kuimawari o↓kUsaNno.
+nyaru/294.wav|2|ʃI↑te ke↑Qka↓tekini ka↓nojoo hi↑doQ↓kU hi↓dokUkizutsUkete ʃi↑ga i↑ma↓ʃIta.
+nyaru/360.wav|2|ra↑ku↓nai ʧa↑yaooQ p a N↓deN.
+nyaru/123.wav|2|na↓tokana i↑Nde↓betsuni a↑no re↑Naigo↓rioʃidyanaikara sa↑a i↑taru↓ʧini ha↑sUkiga ji↑Ngaina.
+nyaru/93.wav|2|n o re↓ga ko↑koni ki↓te ni↑gini da↓sU ku↓mo ʃi↑ma↓ro.
+nyaru/34.wav|2|n o wa↓Qtode ne↑tsu↓ta ta↑naQno ha↓ite a↓rukara ha↑na↓jiNga yo↓ru ra↓na ha↑na↓ri j a i↓N.
+nyaru/248.wav|2|ka↑ʃiina↓a.
+nyaru/394.wav|2|ʧi↓oreto ʧI↑ka↓ine u↑mihaihaihaN↓iN.
+nyaru/148.wav|2|ʧ i.
+nyaru/101.wav|2|no↓aN ʧi↓i a↑ura ka↑waka↓naidakenoo go↓oʃio ʃ I te↑kudasaibu↓ba ta↑oseo su↑re↓ba i↑Qte ko↑to↓wa.
+nyaru/76.wav|2|n a ha↑ʧi↓ta ze↑Nʃi↓ra ji↓no o↑re↓ta ya↑paaiN↓buʃIta a↑ri↓gato.
+nyaru/11.wav|2|n o ma↓ee do↑otaide h i ʃi↓aN ʃI↑tadeNa↓iN.
+nyaru/65.wav|2|do↓riisodesU.
+nyaru/206.wav|2|i↓ma na↑N↓saina no↑kawa ka↓raidakedoto ko↑re↓neN ʃi↓nimo e↓rabenaa ʧi↑gaku↓kookara.
+nyaru/106.wav|2|ka↑zeatai↓toru ta↑gaire↓Naiwa ni↑fa↓atani ju↑u↓teN.
+nyaru/359.wav|2|ha↑naʃi↓dakedo ʧi↓ga sU↑ki↓na hI↑toni o↑ʃIʃokani bu↑ra↓Q ku↑ri↓Q su↑irareta↓ra ma↑ku↓ga.
+nyaru/288.wav|2|ʧo↓Qto ta↑i↓karino a↑saʃa ma↑raa↓aniga to↑ugozaima↓sU.
+nyaru/153.wav|2|u↑tagaini ki↑nara ku↑ne.
+nyaru/203.wav|2|na↑remeQkUsaa↓mani ko↓N ʃi↑naine.
+nyaru/309.wav|2|ga↓nito o↑mo↓Qte ni↑wawa.
+nyaru/383.wav|2|a↑e↓e na↑kaburo↓QkUʧo ri↓biriirinoo ka↑N N do↑buroQka↓kai re↓wa i↑ʧIkama↓i ʃI↑ta ko↑to na↓inonee.
+nyaru/100.wav|2|na↓te mo↑no↓wa pu↑re↓inu ko↑ʃio ʃI↑te ku↑dasa↓i no↑ka ta↑ʧi↓ʃi da↑ʧiN↓joo.
+nyaru/221.wav|2|ʃI↑te ku↑rena mi↑se↓naNni i↑ta i↓ma.
+nyaru/361.wav|2|t e ki↑noo ka↓Nga tsU↑ku↓Qte ho↓N sa↑re ta↓ta.
+nyaru/177.wav|2|ki↓ʃItsu sU↑to o↓meN.

filelists/val.txt

@@ -0,0 +1,18 @@
+wav/nen001_001.wav|0|はい？呼びました？
+wav/nen001_002.wav|0|驚かせたならごめんなさい。通りかかったときにちょうど名前が聞こえてきたので
+wav/nen001_003.wav|0|私に何か用ですか？
+wav/nen001_004.wav|0|いえ、少し気になっただけですから、別に怒っているわけじゃないです。気にしないで下さい
+wav/nen001_005.wav|0|それより仮屋さん、例の件ですが――
+wav/nen001_006.wav|0|喜んでもらえたならなによりです
+wav/nen001_007.wav|0|また何かあったら、いつでも部室に来て下さい
+wav/nen001_008.wav|0|大したことじゃないので。それより体調の方はもういいんですか？
+wav/nen001_009.wav|0|それはよかったです。他に困ったことはありますか？
+wav/nen001_010.wav|0|何かあったらいつでも話して下さい。学院のことじゃなく、私事に関することでも何でも
+wav/nen001_011.wav|0|はい、いつでもどうぞ
+wav/nen001_012.wav|0|保科君も
+wav/nen001_013.wav|0|もし何か困ったことがあれば、力になりますから
+wav/nen001_014.wav|0|そうですか？私には、なにか悩み事があるように見えたりしましたけど……
+wav/nen001_015.wav|0|――なんて、言ってみただけですから、深い意味はありませんよ
+wav/nen001_016.wav|0|いえ、そういうわけじゃなくって……ただ何となく。そんな気がしただけですから
+wav/nen001_017.wav|0|あ、いえ、絶対に秘密というわけじゃないので気にしないで下さい
+wav/nen001_018.wav|0|好奇心だけで来られると困るので、本当に悩んでる人以外には、広めないようにお願いしているだけです

filelists/val.txt.cleaned

@@ -0,0 +1,18 @@
+wav/nen001_001.wav|0|ha↓i? yo↑bima↓ʃIta?
+wav/nen001_002.wav|0|o↑doro↓kasetanara go↑meNnasa↓i. to↑orikaka↓Qta to↓kini ʧo↑odo na↑maega kI↑koete ki↓tanode.
+wav/nen001_003.wav|0|wa↑taʃini na↑nika↓yoodesUka?
+wav/nen001_004.wav|0|i↓e, sU↑ko↓ʃI ki↑ni na↓QtadakedesUkara, be↑tsuni o↑ko↓Qte i↑ru wa↓kejanaidesU. ki↑ni ʃi↑na↓ide ku↑dasa↓i.
+wav/nen001_005.wav|0|so↑reyo↓ri ka↑riyasaN, re↓eno ke↓NdesUga----
+wav/nen001_006.wav|0|yo↑roko↓Nde mo↑raeta↓nara na↓niyoridesU.
+wav/nen001_007.wav|0|ma↑ta na↓nika a↓Qtara, i↓tsudemo bu↑ʃItsuni ki↓te ku↑dasa↓i.
+wav/nen001_008.wav|0|ta↓iʃIta ko↑to↓janainode. so↑reyo↓ri ta↑iʧoono ho↓owa mo↓o i↓i N↓desUka?
+wav/nen001_009.wav|0|so↑rewa yo↓kaQtadesU. ta↓ni ko↑ma↓Qta ko↑to↓wa a↑rima↓sUka?
+wav/nen001_010.wav|0|na↓nika a↓Qtara i↓tsudemo ha↑na↓ʃIte ku↑dasa↓i. ga↑kuiNno ko↑to↓janaku, ʃi↓jini ka↑Nsu↓ru ko↑to↓demo na↓nidemo.
+wav/nen001_011.wav|0|ha↓i, i↓tsudemo do↓ozo.
+wav/nen001_012.wav|0|ho↓ʃinakuNmo.
+wav/nen001_013.wav|0|mo↓ʃi na↓nika ko↑ma↓Qta ko↑to↓ga a↑re↓ba, ʧI↑kara↓ni na↑rima↓sUkara.
+wav/nen001_014.wav|0|so↑odesU↓ka? wa↑taʃiniwa, na↓nika na↑yami↓gotoga a↓ru yo↓oni mi↑eta↓ri ʃi↑ma↓ʃItakedo......
+wav/nen001_015.wav|0|---- na↓Nte, i↑Qte mi↓tadakedesUkara, fU↑ka↓i i↑miwaarimase↓Nyo.
+wav/nen001_016.wav|0|i↓e, so↑oyuu wa↓kejanakuQte...... ta↓da na↑Ntona↓ku. so↑Nna ki↑ga ʃI↑ta↓dakedesUkara.
+wav/nen001_017.wav|0|a, i↓e, ze↑Qtaini hi↑mitsUto i↑u wa↓kejanainode ki↑ni ʃi↑na↓ide ku↑dasa↓i.
+wav/nen001_018.wav|0|ko↑okIʃiNdakede ko↑rare↓ruto ko↑ma↓runode, ho↑Ntooni na↑ya↓Nde ru↑niNi↓gainiwa, hi↑romenai yo↓oni o↑negai ʃI↑te i↑rudakedesU.

filelists/val_filelist.txt

@@ -0,0 +1,6 @@
+louise/VOICE_ID_05733.wav|早く部屋から出てって！着替えはシエスタにしてもらうから。
+louise/VOICE_ID_05734.wav|ぐずぐずしてないで、さっさと出てって！
+louise/VOICE_ID_05735.wav|もう、サイトが早く部屋から出てくれなかったから遅れたんでしょ？
+louise/VOICE_ID_05736.wav|え……えぇ！？あ、いえ、その、なんでもありません。ミスタ・コルベール。
+louise/VOICE_ID_05737.wav|はい。
+louise/VOICE_ID_05738.wav|ま、わたし達には関係ないけどね。

filelists/val_filelist.txt.cleaned

@@ -0,0 +1,6 @@
+louise/VOICE_ID_05733.wav|ha↓yakU he↑ya↓kara de↑teQte! ki↑gaewa ʃi↑e↓sUtani ʃI↑te mo↑rau↓kara.
+louise/VOICE_ID_05734.wav|gu↓zuguzu ʃI↑tena↓ide, sa↓Qsato de↑teQte!
+louise/VOICE_ID_05735.wav|mo↓o, sa↑itoga ha↓yakU he↑ya↓kara de↑te ku↑renakaQta↓kara o↑kureta N↓deʃo?
+louise/VOICE_ID_05736.wav|e…… e↓e!? a, i↓e, so↑no, na↓Ndemo a↑rimase↓N. mi↓sUtako↑rube↓eru.
+louise/VOICE_ID_05737.wav|ha↓i.
+louise/VOICE_ID_05738.wav|m a, wa↑taʃi↓taʧiniwa ka↑Nkee na↓ikedone.

inference.ipynb

@@ -0,0 +1,200 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%matplotlib inline\n",
+    "import matplotlib.pyplot as plt\n",
+    "import IPython.display as ipd\n",
+    "\n",
+    "import os\n",
+    "import json\n",
+    "import math\n",
+    "import torch\n",
+    "from torch import nn\n",
+    "from torch.nn import functional as F\n",
+    "from torch.utils.data import DataLoader\n",
+    "\n",
+    "import commons\n",
+    "import utils\n",
+    "from data_utils import TextAudioLoader, TextAudioCollate, TextAudioSpeakerLoader, TextAudioSpeakerCollate\n",
+    "from models import SynthesizerTrn\n",
+    "from text.symbols import symbols\n",
+    "from text import text_to_sequence\n",
+    "\n",
+    "from scipy.io.wavfile import write\n",
+    "\n",
+    "\n",
+    "def get_text(text, hps):\n",
+    "    text_norm = text_to_sequence(text, hps.data.text_cleaners)\n",
+    "    if hps.data.add_blank:\n",
+    "        text_norm = commons.intersperse(text_norm, 0)\n",
+    "    text_norm = torch.LongTensor(text_norm)\n",
+    "    return text_norm"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## LJ Speech"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "hps = utils.get_hparams_from_file(\"./configs/ljs_base.json\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "net_g = SynthesizerTrn(\n",
+    "    len(symbols),\n",
+    "    hps.data.filter_length // 2 + 1,\n",
+    "    hps.train.segment_size // hps.data.hop_length,\n",
+    "    **hps.model).cuda()\n",
+    "_ = net_g.eval()\n",
+    "\n",
+    "_ = utils.load_checkpoint(\"/path/to/pretrained_ljs.pth\", net_g, None)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "stn_tst = get_text(\"VITS is Awesome!\", hps)\n",
+    "with torch.no_grad():\n",
+    "    x_tst = stn_tst.cuda().unsqueeze(0)\n",
+    "    x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()\n",
+    "    audio = net_g.infer(x_tst, x_tst_lengths, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][0,0].data.cpu().float().numpy()\n",
+    "ipd.display(ipd.Audio(audio, rate=hps.data.sampling_rate, normalize=False))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## VCTK"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "hps = utils.get_hparams_from_file(\"./configs/vctk_base.json\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "net_g = SynthesizerTrn(\n",
+    "    len(symbols),\n",
+    "    hps.data.filter_length // 2 + 1,\n",
+    "    hps.train.segment_size // hps.data.hop_length,\n",
+    "    n_speakers=hps.data.n_speakers,\n",
+    "    **hps.model).cuda()\n",
+    "_ = net_g.eval()\n",
+    "\n",
+    "_ = utils.load_checkpoint(\"/path/to/pretrained_vctk.pth\", net_g, None)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "stn_tst = get_text(\"VITS is Awesome!\", hps)\n",
+    "with torch.no_grad():\n",
+    "    x_tst = stn_tst.cuda().unsqueeze(0)\n",
+    "    x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()\n",
+    "    sid = torch.LongTensor([4]).cuda()\n",
+    "    audio = net_g.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][0,0].data.cpu().float().numpy()\n",
+    "ipd.display(ipd.Audio(audio, rate=hps.data.sampling_rate, normalize=False))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Voice Conversion"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "dataset = TextAudioSpeakerLoader(hps.data.validation_files, hps.data)\n",
+    "collate_fn = TextAudioSpeakerCollate()\n",
+    "loader = DataLoader(dataset, num_workers=8, shuffle=False,\n",
+    "    batch_size=1, pin_memory=True,\n",
+    "    drop_last=True, collate_fn=collate_fn)\n",
+    "data_list = list(loader)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "with torch.no_grad():\n",
+    "    x, x_lengths, spec, spec_lengths, y, y_lengths, sid_src = [x.cuda() for x in data_list[0]]\n",
+    "    sid_tgt1 = torch.LongTensor([1]).cuda()\n",
+    "    sid_tgt2 = torch.LongTensor([2]).cuda()\n",
+    "    sid_tgt3 = torch.LongTensor([4]).cuda()\n",
+    "    audio1 = net_g.voice_conversion(spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_tgt1)[0][0,0].data.cpu().float().numpy()\n",
+    "    audio2 = net_g.voice_conversion(spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_tgt2)[0][0,0].data.cpu().float().numpy()\n",
+    "    audio3 = net_g.voice_conversion(spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_tgt3)[0][0,0].data.cpu().float().numpy()\n",
+    "print(\"Original SID: %d\" % sid_src.item())\n",
+    "ipd.display(ipd.Audio(y[0].cpu().numpy(), rate=hps.data.sampling_rate, normalize=False))\n",
+    "print(\"Converted SID: %d\" % sid_tgt1.item())\n",
+    "ipd.display(ipd.Audio(audio1, rate=hps.data.sampling_rate, normalize=False))\n",
+    "print(\"Converted SID: %d\" % sid_tgt2.item())\n",
+    "ipd.display(ipd.Audio(audio2, rate=hps.data.sampling_rate, normalize=False))\n",
+    "print(\"Converted SID: %d\" % sid_tgt3.item())\n",
+    "ipd.display(ipd.Audio(audio3, rate=hps.data.sampling_rate, normalize=False))"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.7.7"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

log.log

@@ -0,0 +1,146 @@
+Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
+Collecting Cython==0.29.21
+  Downloading Cython-0.29.21-cp37-cp37m-manylinux1_x86_64.whl (2.0 MB)
+Collecting librosa==0.8.0
+  Downloading librosa-0.8.0.tar.gz (183 kB)
+Collecting matplotlib==3.3.1
+  Downloading matplotlib-3.3.1-cp37-cp37m-manylinux1_x86_64.whl (11.6 MB)
+Requirement already satisfied: numpy==1.21.6 in /usr/local/lib/python3.7/dist-packages (from -r requirements.txt (line 4)) (1.21.6)
+Collecting phonemizer==2.2.1
+  Downloading phonemizer-2.2.1-py3-none-any.whl (49 kB)
+Collecting scipy==1.5.2
+  Downloading scipy-1.5.2-cp37-cp37m-manylinux1_x86_64.whl (25.9 MB)
+Collecting tensorboard==2.3.0
+  Downloading tensorboard-2.3.0-py3-none-any.whl (6.8 MB)
+Collecting torch==1.6.0
+  Downloading torch-1.6.0-cp37-cp37m-manylinux1_x86_64.whl (748.8 MB)
+Collecting torchvision==0.7.0
+  Downloading torchvision-0.7.0-cp37-cp37m-manylinux1_x86_64.whl (5.9 MB)
+Collecting Unidecode==1.1.1
+  Downloading Unidecode-1.1.1-py2.py3-none-any.whl (238 kB)
+Collecting pyopenjtalk==0.2.0
+  Downloading pyopenjtalk-0.2.0.tar.gz (1.5 MB)
+  Installing build dependencies: started
+  Installing build dependencies: finished with status 'done'
+  Getting requirements to build wheel: started
+  Getting requirements to build wheel: finished with status 'done'
+    Preparing wheel metadata: started
+    Preparing wheel metadata: finished with status 'done'
+Collecting jamo==0.4.1
+  Downloading jamo-0.4.1-py3-none-any.whl (9.5 kB)
+Collecting pypinyin==0.44.0
+  Downloading pypinyin-0.44.0-py2.py3-none-any.whl (1.3 MB)
+Requirement already satisfied: jieba==0.42.1 in /usr/local/lib/python3.7/dist-packages (from -r requirements.txt (line 14)) (0.42.1)
+Requirement already satisfied: audioread>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa==0.8.0->-r requirements.txt (line 2)) (2.1.9)
+Requirement already satisfied: scikit-learn!=0.19.0,>=0.14.0 in /usr/local/lib/python3.7/dist-packages (from librosa==0.8.0->-r requirements.txt (line 2)) (1.0.2)
+Requirement already satisfied: joblib>=0.14 in /usr/local/lib/python3.7/dist-packages (from librosa==0.8.0->-r requirements.txt (line 2)) (1.1.0)
+Requirement already satisfied: decorator>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa==0.8.0->-r requirements.txt (line 2)) (4.4.2)
+Requirement already satisfied: resampy>=0.2.2 in /usr/local/lib/python3.7/dist-packages (from librosa==0.8.0->-r requirements.txt (line 2)) (0.4.0)
+Requirement already satisfied: numba>=0.43.0 in /usr/local/lib/python3.7/dist-packages (from librosa==0.8.0->-r requirements.txt (line 2)) (0.56.0)
+Requirement already satisfied: soundfile>=0.9.0 in /usr/local/lib/python3.7/dist-packages (from librosa==0.8.0->-r requirements.txt (line 2)) (0.10.3.post1)
+Requirement already satisfied: pooch>=1.0 in /usr/local/lib/python3.7/dist-packages (from librosa==0.8.0->-r requirements.txt (line 2)) (1.6.0)
+Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.3 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.3.1->-r requirements.txt (line 3)) (3.0.9)
+Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.3.1->-r requirements.txt (line 3)) (0.11.0)
+Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.3.1->-r requirements.txt (line 3)) (2.8.2)
+Requirement already satisfied: pillow>=6.2.0 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.3.1->-r requirements.txt (line 3)) (7.1.2)
+Requirement already satisfied: certifi>=2020.06.20 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.3.1->-r requirements.txt (line 3)) (2022.6.15)
+Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib==3.3.1->-r requirements.txt (line 3)) (1.4.4)
+Collecting segments
+  Downloading segments-2.2.1-py2.py3-none-any.whl (15 kB)
+Requirement already satisfied: attrs>=18.1 in /usr/local/lib/python3.7/dist-packages (from phonemizer==2.2.1->-r requirements.txt (line 5)) (22.1.0)
+Requirement already satisfied: google-auth<2,>=1.6.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (1.35.0)
+Requirement already satisfied: setuptools>=41.0.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (57.4.0)
+Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (1.8.1)
+Requirement already satisfied: wheel>=0.26 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (0.37.1)
+Requirement already satisfied: grpcio>=1.24.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (1.47.0)
+Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (0.4.6)
+Requirement already satisfied: requests<3,>=2.21.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (2.23.0)
+Requirement already satisfied: protobuf>=3.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (3.17.3)
+Requirement already satisfied: absl-py>=0.4 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (1.2.0)
+Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (3.4.1)
+Requirement already satisfied: six>=1.10.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (1.15.0)
+Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.7/dist-packages (from tensorboard==2.3.0->-r requirements.txt (line 7)) (1.0.1)
+Requirement already satisfied: future in /usr/local/lib/python3.7/dist-packages (from torch==1.6.0->-r requirements.txt (line 8)) (0.16.0)
+Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from pyopenjtalk==0.2.0->-r requirements.txt (line 11)) (4.64.0)
+Requirement already satisfied: rsa<5,>=3.1.4 in /usr/local/lib/python3.7/dist-packages (from google-auth<2,>=1.6.3->tensorboard==2.3.0->-r requirements.txt (line 7)) (4.9)
+Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from google-auth<2,>=1.6.3->tensorboard==2.3.0->-r requirements.txt (line 7)) (4.2.4)
+Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from google-auth<2,>=1.6.3->tensorboard==2.3.0->-r requirements.txt (line 7)) (0.2.8)
+Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.7/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard==2.3.0->-r requirements.txt (line 7)) (1.3.1)
+Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from kiwisolver>=1.0.1->matplotlib==3.3.1->-r requirements.txt (line 3)) (4.1.1)
+Requirement already satisfied: importlib-metadata>=4.4 in /usr/local/lib/python3.7/dist-packages (from markdown>=2.6.8->tensorboard==2.3.0->-r requirements.txt (line 7)) (4.12.0)
+Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata>=4.4->markdown>=2.6.8->tensorboard==2.3.0->-r requirements.txt (line 7)) (3.8.1)
+Requirement already satisfied: llvmlite<0.40,>=0.39.0dev0 in /usr/local/lib/python3.7/dist-packages (from numba>=0.43.0->librosa==0.8.0->-r requirements.txt (line 2)) (0.39.0)
+Requirement already satisfied: appdirs>=1.3.0 in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa==0.8.0->-r requirements.txt (line 2)) (1.4.4)
+Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa==0.8.0->-r requirements.txt (line 2)) (21.3)
+Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.7/dist-packages (from pyasn1-modules>=0.2.1->google-auth<2,>=1.6.3->tensorboard==2.3.0->-r requirements.txt (line 7)) (0.4.8)
+Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard==2.3.0->-r requirements.txt (line 7)) (1.24.3)
+Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard==2.3.0->-r requirements.txt (line 7)) (3.0.4)
+Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests<3,>=2.21.0->tensorboard==2.3.0->-r requirements.txt (line 7)) (2.10)
+Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard==2.3.0->-r requirements.txt (line 7)) (3.2.0)
+Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn!=0.19.0,>=0.14.0->librosa==0.8.0->-r requirements.txt (line 2)) (3.1.0)
+Requirement already satisfied: cffi>=1.0 in /usr/local/lib/python3.7/dist-packages (from soundfile>=0.9.0->librosa==0.8.0->-r requirements.txt (line 2)) (1.15.1)
+Requirement already satisfied: pycparser in /usr/local/lib/python3.7/dist-packages (from cffi>=1.0->soundfile>=0.9.0->librosa==0.8.0->-r requirements.txt (line 2)) (2.21)
+Collecting csvw>=1.5.6
+  Downloading csvw-3.1.1-py2.py3-none-any.whl (56 kB)
+Requirement already satisfied: regex in /usr/local/lib/python3.7/dist-packages (from segments->phonemizer==2.2.1->-r requirements.txt (line 5)) (2022.6.2)
+Collecting clldutils>=1.7.3
+  Downloading clldutils-3.12.0-py2.py3-none-any.whl (197 kB)
+Requirement already satisfied: tabulate>=0.7.7 in /usr/local/lib/python3.7/dist-packages (from clldutils>=1.7.3->segments->phonemizer==2.2.1->-r requirements.txt (line 5)) (0.8.10)
+Collecting colorlog
+  Downloading colorlog-6.6.0-py2.py3-none-any.whl (11 kB)
+Requirement already satisfied: babel in /usr/local/lib/python3.7/dist-packages (from csvw>=1.5.6->segments->phonemizer==2.2.1->-r requirements.txt (line 5)) (2.10.3)
+Collecting rdflib
+  Downloading rdflib-6.2.0-py3-none-any.whl (500 kB)
+Requirement already satisfied: jsonschema in /usr/local/lib/python3.7/dist-packages (from csvw>=1.5.6->segments->phonemizer==2.2.1->-r requirements.txt (line 5)) (4.3.3)
+Collecting colorama
+  Downloading colorama-0.4.5-py2.py3-none-any.whl (16 kB)
+Collecting rfc3986<2
+  Downloading rfc3986-1.5.0-py2.py3-none-any.whl (31 kB)
+Collecting language-tags
+  Downloading language_tags-1.1.0-py2.py3-none-any.whl (210 kB)
+Collecting isodate
+  Downloading isodate-0.6.1-py2.py3-none-any.whl (41 kB)
+Requirement already satisfied: uritemplate>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from csvw>=1.5.6->segments->phonemizer==2.2.1->-r requirements.txt (line 5)) (3.0.1)
+Requirement already satisfied: pytz>=2015.7 in /usr/local/lib/python3.7/dist-packages (from babel->csvw>=1.5.6->segments->phonemizer==2.2.1->-r requirements.txt (line 5)) (2022.1)
+Requirement already satisfied: pyrsistent!=0.17.0,!=0.17.1,!=0.17.2,>=0.14.0 in /usr/local/lib/python3.7/dist-packages (from jsonschema->csvw>=1.5.6->segments->phonemizer==2.2.1->-r requirements.txt (line 5)) (0.18.1)
+Requirement already satisfied: importlib-resources>=1.4.0 in /usr/local/lib/python3.7/dist-packages (from jsonschema->csvw>=1.5.6->segments->phonemizer==2.2.1->-r requirements.txt (line 5)) (5.9.0)
+Building wheels for collected packages: librosa, pyopenjtalk
+  Building wheel for librosa (setup.py): started
+  Building wheel for librosa (setup.py): finished with status 'done'
+  Created wheel for librosa: filename=librosa-0.8.0-py3-none-any.whl size=201396 sha256=69a746a2373b77774c1b66e31e7eba0bfedeb1d18e378aa9180fd3f7d4019e57
+  Stored in directory: /root/.cache/pip/wheels/de/1e/aa/d91797ae7e1ce11853ee100bee9d1781ae9d750e7458c95afb
+  Building wheel for pyopenjtalk (PEP 517): started
+  Building wheel for pyopenjtalk (PEP 517): finished with status 'done'
+  Created wheel for pyopenjtalk: filename=pyopenjtalk-0.2.0-cp37-cp37m-linux_x86_64.whl size=4431836 sha256=68551b95c2c9065b6654c4e04e0e1631c14d9149046836224effdb990ad77f71
+  Stored in directory: /root/.cache/pip/wheels/10/56/0e/435dc1aec0d8614a489abfc51da4fd54ff6e8b33bf978f2081
+Successfully built librosa pyopenjtalk
+Installing collected packages: isodate, rfc3986, rdflib, language-tags, colorama, csvw, colorlog, scipy, clldutils, torch, segments, Cython, Unidecode, torchvision, tensorboard, pypinyin, pyopenjtalk, phonemizer, matplotlib, librosa, jamo
+  Attempting uninstall: scipy
+    Found existing installation: scipy 1.7.3
+    Uninstalling scipy-1.7.3:
+      Successfully uninstalled scipy-1.7.3
+  Attempting uninstall: torch
+    Found existing installation: torch 1.12.1+cu113
+    Uninstalling torch-1.12.1+cu113:
+      Successfully uninstalled torch-1.12.1+cu113
+  Attempting uninstall: Cython
+    Found existing installation: Cython 0.29.32
+    Uninstalling Cython-0.29.32:
+      Successfully uninstalled Cython-0.29.32
+  Attempting uninstall: torchvision
+    Found existing installation: torchvision 0.13.1+cu113
+    Uninstalling torchvision-0.13.1+cu113:
+      Successfully uninstalled torchvision-0.13.1+cu113
+  Attempting uninstall: tensorboard
+    Found existing installation: tensorboard 2.8.0
+    Uninstalling tensorboard-2.8.0:
+      Successfully uninstalled tensorboard-2.8.0
+  Attempting uninstall: matplotlib
+    Found existing installation: matplotlib 3.2.2
+    Uninstalling matplotlib-3.2.2:
+      Successfully uninstalled matplotlib-3.2.2
+  Attempting uninstall: librosa
+    Found existing installation: librosa 0.8.1
+    Uninstalling librosa-0.8.1:
+      Successfully uninstalled librosa-0.8.1
+Successfully installed Cython-0.29.21 Unidecode-1.1.1 clldutils-3.12.0 colorama-0.4.5 colorlog-6.6.0 csvw-3.1.1 isodate-0.6.1 jamo-0.4.1 language-tags-1.1.0 librosa-0.8.0 matplotlib-3.3.1 phonemizer-2.2.1 pyopenjtalk-0.2.0 pypinyin-0.44.0 rdflib-6.2.0 rfc3986-1.5.0 scipy-1.5.2 segments-2.2.1 tensorboard-2.3.0 torch-1.6.0 torchvision-0.7.0

losses.py

@@ -0,0 +1,61 @@
+import torch 
+from torch.nn import functional as F
+
+import commons
+
+
+def feature_loss(fmap_r, fmap_g):
+  loss = 0
+  for dr, dg in zip(fmap_r, fmap_g):
+    for rl, gl in zip(dr, dg):
+      rl = rl.float().detach()
+      gl = gl.float()
+      loss += torch.mean(torch.abs(rl - gl))
+
+  return loss * 2 
+
+
+def discriminator_loss(disc_real_outputs, disc_generated_outputs):
+  loss = 0
+  r_losses = []
+  g_losses = []
+  for dr, dg in zip(disc_real_outputs, disc_generated_outputs):
+    dr = dr.float()
+    dg = dg.float()
+    r_loss = torch.mean((1-dr)**2)
+    g_loss = torch.mean(dg**2)
+    loss += (r_loss + g_loss)
+    r_losses.append(r_loss.item())
+    g_losses.append(g_loss.item())
+
+  return loss, r_losses, g_losses
+
+
+def generator_loss(disc_outputs):
+  loss = 0
+  gen_losses = []
+  for dg in disc_outputs:
+    dg = dg.float()
+    l = torch.mean((1-dg)**2)
+    gen_losses.append(l)
+    loss += l
+
+  return loss, gen_losses
+
+
+def kl_loss(z_p, logs_q, m_p, logs_p, z_mask):
+  """
+  z_p, logs_q: [b, h, t_t]
+  m_p, logs_p: [b, h, t_t]
+  """
+  z_p = z_p.float()
+  logs_q = logs_q.float()
+  m_p = m_p.float()
+  logs_p = logs_p.float()
+  z_mask = z_mask.float()
+
+  kl = logs_p - logs_q - 0.5
+  kl += 0.5 * ((z_p - m_p)**2) * torch.exp(-2. * logs_p)
+  kl = torch.sum(kl * z_mask)
+  l = kl / torch.sum(z_mask)
+  return l

mel_processing.py

@@ -0,0 +1,112 @@
+import math
+import os
+import random
+import torch
+from torch import nn
+import torch.nn.functional as F
+import torch.utils.data
+import numpy as np
+import librosa
+import librosa.util as librosa_util
+from librosa.util import normalize, pad_center, tiny
+from scipy.signal import get_window
+from scipy.io.wavfile import read
+from librosa.filters import mel as librosa_mel_fn
+
+MAX_WAV_VALUE = 32768.0
+
+
+def dynamic_range_compression_torch(x, C=1, clip_val=1e-5):
+    """
+    PARAMS
+    ------
+    C: compression factor
+    """
+    return torch.log(torch.clamp(x, min=clip_val) * C)
+
+
+def dynamic_range_decompression_torch(x, C=1):
+    """
+    PARAMS
+    ------
+    C: compression factor used to compress
+    """
+    return torch.exp(x) / C
+
+
+def spectral_normalize_torch(magnitudes):
+    output = dynamic_range_compression_torch(magnitudes)
+    return output
+
+
+def spectral_de_normalize_torch(magnitudes):
+    output = dynamic_range_decompression_torch(magnitudes)
+    return output
+
+
+mel_basis = {}
+hann_window = {}
+
+
+def spectrogram_torch(y, n_fft, sampling_rate, hop_size, win_size, center=False):
+    if torch.min(y) < -1.:
+        print('min value is ', torch.min(y))
+    if torch.max(y) > 1.:
+        print('max value is ', torch.max(y))
+
+    global hann_window
+    dtype_device = str(y.dtype) + '_' + str(y.device)
+    wnsize_dtype_device = str(win_size) + '_' + dtype_device
+    if wnsize_dtype_device not in hann_window:
+        hann_window[wnsize_dtype_device] = torch.hann_window(win_size).to(dtype=y.dtype, device=y.device)
+
+    y = torch.nn.functional.pad(y.unsqueeze(1), (int((n_fft-hop_size)/2), int((n_fft-hop_size)/2)), mode='reflect')
+    y = y.squeeze(1)
+
+    spec = torch.stft(y, n_fft, hop_length=hop_size, win_length=win_size, window=hann_window[wnsize_dtype_device],
+                      center=center, pad_mode='reflect', normalized=False, onesided=True)
+
+    spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)
+    return spec
+
+
+def spec_to_mel_torch(spec, n_fft, num_mels, sampling_rate, fmin, fmax):
+    global mel_basis
+    dtype_device = str(spec.dtype) + '_' + str(spec.device)
+    fmax_dtype_device = str(fmax) + '_' + dtype_device
+    if fmax_dtype_device not in mel_basis:
+        mel = librosa_mel_fn(sampling_rate, n_fft, num_mels, fmin, fmax)
+        mel_basis[fmax_dtype_device] = torch.from_numpy(mel).to(dtype=spec.dtype, device=spec.device)
+    spec = torch.matmul(mel_basis[fmax_dtype_device], spec)
+    spec = spectral_normalize_torch(spec)
+    return spec
+
+
+def mel_spectrogram_torch(y, n_fft, num_mels, sampling_rate, hop_size, win_size, fmin, fmax, center=False):
+    if torch.min(y) < -1.:
+        print('min value is ', torch.min(y))
+    if torch.max(y) > 1.:
+        print('max value is ', torch.max(y))
+
+    global mel_basis, hann_window
+    dtype_device = str(y.dtype) + '_' + str(y.device)
+    fmax_dtype_device = str(fmax) + '_' + dtype_device
+    wnsize_dtype_device = str(win_size) + '_' + dtype_device
+    if fmax_dtype_device not in mel_basis:
+        mel = librosa_mel_fn(sampling_rate, n_fft, num_mels, fmin, fmax)
+        mel_basis[fmax_dtype_device] = torch.from_numpy(mel).to(dtype=y.dtype, device=y.device)
+    if wnsize_dtype_device not in hann_window:
+        hann_window[wnsize_dtype_device] = torch.hann_window(win_size).to(dtype=y.dtype, device=y.device)
+
+    y = torch.nn.functional.pad(y.unsqueeze(1), (int((n_fft-hop_size)/2), int((n_fft-hop_size)/2)), mode='reflect')
+    y = y.squeeze(1)
+
+    spec = torch.stft(y, n_fft, hop_length=hop_size, win_length=win_size, window=hann_window[wnsize_dtype_device],
+                      center=center, pad_mode='reflect', normalized=False, onesided=True)
+
+    spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)
+
+    spec = torch.matmul(mel_basis[fmax_dtype_device], spec)
+    spec = spectral_normalize_torch(spec)
+
+    return spec

models.py

@@ -0,0 +1,535 @@
+import copy
+import math
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+import commons
+import modules
+import attentions
+import monotonic_align
+
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm, spectral_norm
+from commons import init_weights, get_padding
+
+
+class StochasticDurationPredictor(nn.Module):
+  def __init__(self, in_channels, filter_channels, kernel_size, p_dropout, n_flows=4, gin_channels=0):
+    super().__init__()
+    filter_channels = in_channels # it needs to be removed from future version.
+    self.in_channels = in_channels
+    self.filter_channels = filter_channels
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.n_flows = n_flows
+    self.gin_channels = gin_channels
+
+    self.log_flow = modules.Log()
+    self.flows = nn.ModuleList()
+    self.flows.append(modules.ElementwiseAffine(2))
+    for i in range(n_flows):
+      self.flows.append(modules.ConvFlow(2, filter_channels, kernel_size, n_layers=3))
+      self.flows.append(modules.Flip())
+
+    self.post_pre = nn.Conv1d(1, filter_channels, 1)
+    self.post_proj = nn.Conv1d(filter_channels, filter_channels, 1)
+    self.post_convs = modules.DDSConv(filter_channels, kernel_size, n_layers=3, p_dropout=p_dropout)
+    self.post_flows = nn.ModuleList()
+    self.post_flows.append(modules.ElementwiseAffine(2))
+    for i in range(4):
+      self.post_flows.append(modules.ConvFlow(2, filter_channels, kernel_size, n_layers=3))
+      self.post_flows.append(modules.Flip())
+
+    self.pre = nn.Conv1d(in_channels, filter_channels, 1)
+    self.proj = nn.Conv1d(filter_channels, filter_channels, 1)
+    self.convs = modules.DDSConv(filter_channels, kernel_size, n_layers=3, p_dropout=p_dropout)
+    if gin_channels != 0:
+      self.cond = nn.Conv1d(gin_channels, filter_channels, 1)
+
+  def forward(self, x, x_mask, w=None, g=None, reverse=False, noise_scale=1.0):
+    x = torch.detach(x)
+    x = self.pre(x)
+    if g is not None:
+      g = torch.detach(g)
+      x = x + self.cond(g)
+    x = self.convs(x, x_mask)
+    x = self.proj(x) * x_mask
+
+    if not reverse:
+      flows = self.flows
+      assert w is not None
+
+      logdet_tot_q = 0 
+      h_w = self.post_pre(w)
+      h_w = self.post_convs(h_w, x_mask)
+      h_w = self.post_proj(h_w) * x_mask
+      e_q = torch.randn(w.size(0), 2, w.size(2)).to(device=x.device, dtype=x.dtype) * x_mask
+      z_q = e_q
+      for flow in self.post_flows:
+        z_q, logdet_q = flow(z_q, x_mask, g=(x + h_w))
+        logdet_tot_q += logdet_q
+      z_u, z1 = torch.split(z_q, [1, 1], 1) 
+      u = torch.sigmoid(z_u) * x_mask
+      z0 = (w - u) * x_mask
+      logdet_tot_q += torch.sum((F.logsigmoid(z_u) + F.logsigmoid(-z_u)) * x_mask, [1,2])
+      logq = torch.sum(-0.5 * (math.log(2*math.pi) + (e_q**2)) * x_mask, [1,2]) - logdet_tot_q
+
+      logdet_tot = 0
+      z0, logdet = self.log_flow(z0, x_mask)
+      logdet_tot += logdet
+      z = torch.cat([z0, z1], 1)
+      for flow in flows:
+        z, logdet = flow(z, x_mask, g=x, reverse=reverse)
+        logdet_tot = logdet_tot + logdet
+      nll = torch.sum(0.5 * (math.log(2*math.pi) + (z**2)) * x_mask, [1,2]) - logdet_tot
+      return nll + logq # [b]
+    else:
+      flows = list(reversed(self.flows))
+      flows = flows[:-2] + [flows[-1]] # remove a useless vflow
+      z = torch.randn(x.size(0), 2, x.size(2)).to(device=x.device, dtype=x.dtype) * noise_scale
+      for flow in flows:
+        z = flow(z, x_mask, g=x, reverse=reverse)
+      z0, z1 = torch.split(z, [1, 1], 1)
+      logw = z0
+      return logw
+
+
+class DurationPredictor(nn.Module):
+  def __init__(self, in_channels, filter_channels, kernel_size, p_dropout, gin_channels=0):
+    super().__init__()
+
+    self.in_channels = in_channels
+    self.filter_channels = filter_channels
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.gin_channels = gin_channels
+
+    self.drop = nn.Dropout(p_dropout)
+    self.conv_1 = nn.Conv1d(in_channels, filter_channels, kernel_size, padding=kernel_size//2)
+    self.norm_1 = modules.LayerNorm(filter_channels)
+    self.conv_2 = nn.Conv1d(filter_channels, filter_channels, kernel_size, padding=kernel_size//2)
+    self.norm_2 = modules.LayerNorm(filter_channels)
+    self.proj = nn.Conv1d(filter_channels, 1, 1)
+
+    if gin_channels != 0:
+      self.cond = nn.Conv1d(gin_channels, in_channels, 1)
+
+  def forward(self, x, x_mask, g=None):
+    x = torch.detach(x)
+    if g is not None:
+      g = torch.detach(g)
+      x = x + self.cond(g)
+    x = self.conv_1(x * x_mask)
+    x = torch.relu(x)
+    x = self.norm_1(x)
+    x = self.drop(x)
+    x = self.conv_2(x * x_mask)
+    x = torch.relu(x)
+    x = self.norm_2(x)
+    x = self.drop(x)
+    x = self.proj(x * x_mask)
+    return x * x_mask
+
+
+class TextEncoder(nn.Module):
+  def __init__(self,
+      n_vocab,
+      out_channels,
+      hidden_channels,
+      filter_channels,
+      n_heads,
+      n_layers,
+      kernel_size,
+      p_dropout):
+    super().__init__()
+    self.n_vocab = n_vocab
+    self.out_channels = out_channels
+    self.hidden_channels = hidden_channels
+    self.filter_channels = filter_channels
+    self.n_heads = n_heads
+    self.n_layers = n_layers
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+
+    self.emb = nn.Embedding(n_vocab, hidden_channels)
+    nn.init.normal_(self.emb.weight, 0.0, hidden_channels**-0.5)
+
+    self.encoder = attentions.Encoder(
+      hidden_channels,
+      filter_channels,
+      n_heads,
+      n_layers,
+      kernel_size,
+      p_dropout)
+    self.proj= nn.Conv1d(hidden_channels, out_channels * 2, 1)
+
+  def forward(self, x, x_lengths):
+    x = self.emb(x) * math.sqrt(self.hidden_channels) # [b, t, h]
+    # print(x.shape)
+    x = torch.transpose(x, 1, -1) # [b, h, t]
+    x_mask = torch.unsqueeze(commons.sequence_mask(x_lengths, x.size(2)), 1).to(x.dtype)
+
+    x = self.encoder(x * x_mask, x_mask)
+    stats = self.proj(x) * x_mask
+
+    m, logs = torch.split(stats, self.out_channels, dim=1)
+    return x, m, logs, x_mask
+
+
+class ResidualCouplingBlock(nn.Module):
+  def __init__(self,
+      channels,
+      hidden_channels,
+      kernel_size,
+      dilation_rate,
+      n_layers,
+      n_flows=4,
+      gin_channels=0):
+    super().__init__()
+    self.channels = channels
+    self.hidden_channels = hidden_channels
+    self.kernel_size = kernel_size
+    self.dilation_rate = dilation_rate
+    self.n_layers = n_layers
+    self.n_flows = n_flows
+    self.gin_channels = gin_channels
+
+    self.flows = nn.ModuleList()
+    for i in range(n_flows):
+      self.flows.append(modules.ResidualCouplingLayer(channels, hidden_channels, kernel_size, dilation_rate, n_layers, gin_channels=gin_channels, mean_only=True))
+      self.flows.append(modules.Flip())
+
+  def forward(self, x, x_mask, g=None, reverse=False):
+    if not reverse:
+      for flow in self.flows:
+        x, _ = flow(x, x_mask, g=g, reverse=reverse)
+    else:
+      for flow in reversed(self.flows):
+        x = flow(x, x_mask, g=g, reverse=reverse)
+    return x
+
+
+class PosteriorEncoder(nn.Module):
+  def __init__(self,
+      in_channels,
+      out_channels,
+      hidden_channels,
+      kernel_size,
+      dilation_rate,
+      n_layers,
+      gin_channels=0):
+    super().__init__()
+    self.in_channels = in_channels
+    self.out_channels = out_channels
+    self.hidden_channels = hidden_channels
+    self.kernel_size = kernel_size
+    self.dilation_rate = dilation_rate
+    self.n_layers = n_layers
+    self.gin_channels = gin_channels
+
+    self.pre = nn.Conv1d(in_channels, hidden_channels, 1)
+    self.enc = modules.WN(hidden_channels, kernel_size, dilation_rate, n_layers, gin_channels=gin_channels)
+    self.proj = nn.Conv1d(hidden_channels, out_channels * 2, 1)
+
+  def forward(self, x, x_lengths, g=None):
+    x_mask = torch.unsqueeze(commons.sequence_mask(x_lengths, x.size(2)), 1).to(x.dtype)
+    x = self.pre(x) * x_mask
+    x = self.enc(x, x_mask, g=g)
+    stats = self.proj(x) * x_mask
+    m, logs = torch.split(stats, self.out_channels, dim=1)
+    z = (m + torch.randn_like(m) * torch.exp(logs)) * x_mask
+    return z, m, logs, x_mask
+
+
+class Generator(torch.nn.Module):
+    def __init__(self, initial_channel, resblock, resblock_kernel_sizes, resblock_dilation_sizes, upsample_rates, upsample_initial_channel, upsample_kernel_sizes, gin_channels=0):
+        super(Generator, self).__init__()
+        self.num_kernels = len(resblock_kernel_sizes)
+        self.num_upsamples = len(upsample_rates)
+        self.conv_pre = Conv1d(initial_channel, upsample_initial_channel, 7, 1, padding=3)
+        resblock = modules.ResBlock1 if resblock == '1' else modules.ResBlock2
+
+        self.ups = nn.ModuleList()
+        for i, (u, k) in enumerate(zip(upsample_rates, upsample_kernel_sizes)):
+            self.ups.append(weight_norm(
+                ConvTranspose1d(upsample_initial_channel//(2**i), upsample_initial_channel//(2**(i+1)),
+                                k, u, padding=(k-u)//2)))
+
+        self.resblocks = nn.ModuleList()
+        for i in range(len(self.ups)):
+            ch = upsample_initial_channel//(2**(i+1))
+            for j, (k, d) in enumerate(zip(resblock_kernel_sizes, resblock_dilation_sizes)):
+                self.resblocks.append(resblock(ch, k, d))
+
+        self.conv_post = Conv1d(ch, 1, 7, 1, padding=3, bias=False)
+        self.ups.apply(init_weights)
+
+        if gin_channels != 0:
+            self.cond = nn.Conv1d(gin_channels, upsample_initial_channel, 1)
+
+    def forward(self, x, g=None):
+        x = self.conv_pre(x)
+        if g is not None:
+          x = x + self.cond(g)
+
+        for i in range(self.num_upsamples):
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            x = self.ups[i](x)
+            xs = None
+            for j in range(self.num_kernels):
+                if xs is None:
+                    xs = self.resblocks[i*self.num_kernels+j](x)
+                else:
+                    xs += self.resblocks[i*self.num_kernels+j](x)
+            x = xs / self.num_kernels
+        x = F.leaky_relu(x)
+        x = self.conv_post(x)
+        x = torch.tanh(x)
+
+        return x
+
+    def remove_weight_norm(self):
+        print('Removing weight norm...')
+        for l in self.ups:
+            remove_weight_norm(l)
+        for l in self.resblocks:
+            l.remove_weight_norm()
+
+
+class DiscriminatorP(torch.nn.Module):
+    def __init__(self, period, kernel_size=5, stride=3, use_spectral_norm=False):
+        super(DiscriminatorP, self).__init__()
+        self.period = period
+        self.use_spectral_norm = use_spectral_norm
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList([
+            norm_f(Conv2d(1, 32, (kernel_size, 1), (stride, 1), padding=(get_padding(kernel_size, 1), 0))),
+            norm_f(Conv2d(32, 128, (kernel_size, 1), (stride, 1), padding=(get_padding(kernel_size, 1), 0))),
+            norm_f(Conv2d(128, 512, (kernel_size, 1), (stride, 1), padding=(get_padding(kernel_size, 1), 0))),
+            norm_f(Conv2d(512, 1024, (kernel_size, 1), (stride, 1), padding=(get_padding(kernel_size, 1), 0))),
+            norm_f(Conv2d(1024, 1024, (kernel_size, 1), 1, padding=(get_padding(kernel_size, 1), 0))),
+        ])
+        self.conv_post = norm_f(Conv2d(1024, 1, (3, 1), 1, padding=(1, 0)))
+
+    def forward(self, x):
+        fmap = []
+
+        # 1d to 2d
+        b, c, t = x.shape
+        if t % self.period != 0: # pad first
+            n_pad = self.period - (t % self.period)
+            x = F.pad(x, (0, n_pad), "reflect")
+            t = t + n_pad
+        x = x.view(b, c, t // self.period, self.period)
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
+
+
+class DiscriminatorS(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(DiscriminatorS, self).__init__()
+        norm_f = weight_norm if use_spectral_norm == False else spectral_norm
+        self.convs = nn.ModuleList([
+            norm_f(Conv1d(1, 16, 15, 1, padding=7)),
+            norm_f(Conv1d(16, 64, 41, 4, groups=4, padding=20)),
+            norm_f(Conv1d(64, 256, 41, 4, groups=16, padding=20)),
+            norm_f(Conv1d(256, 1024, 41, 4, groups=64, padding=20)),
+            norm_f(Conv1d(1024, 1024, 41, 4, groups=256, padding=20)),
+            norm_f(Conv1d(1024, 1024, 5, 1, padding=2)),
+        ])
+        self.conv_post = norm_f(Conv1d(1024, 1, 3, 1, padding=1))
+
+    def forward(self, x):
+        fmap = []
+
+        for l in self.convs:
+            x = l(x)
+            x = F.leaky_relu(x, modules.LRELU_SLOPE)
+            fmap.append(x)
+        x = self.conv_post(x)
+        fmap.append(x)
+        x = torch.flatten(x, 1, -1)
+
+        return x, fmap
+
+
+class MultiPeriodDiscriminator(torch.nn.Module):
+    def __init__(self, use_spectral_norm=False):
+        super(MultiPeriodDiscriminator, self).__init__()
+        periods = [2,3,5,7,11]
+
+        discs = [DiscriminatorS(use_spectral_norm=use_spectral_norm)]
+        discs = discs + [DiscriminatorP(i, use_spectral_norm=use_spectral_norm) for i in periods]
+        self.discriminators = nn.ModuleList(discs)
+
+    def forward(self, y, y_hat):
+        y_d_rs = []
+        y_d_gs = []
+        fmap_rs = []
+        fmap_gs = []
+        for i, d in enumerate(self.discriminators):
+            y_d_r, fmap_r = d(y)
+            y_d_g, fmap_g = d(y_hat)
+            y_d_rs.append(y_d_r)
+            y_d_gs.append(y_d_g)
+            fmap_rs.append(fmap_r)
+            fmap_gs.append(fmap_g)
+
+        return y_d_rs, y_d_gs, fmap_rs, fmap_gs
+
+
+
+class SynthesizerTrn(nn.Module):
+  """
+  Synthesizer for Training
+  """
+
+  def __init__(self, 
+    n_vocab,
+    spec_channels,
+    segment_size,
+    inter_channels,
+    hidden_channels,
+    filter_channels,
+    n_heads,
+    n_layers,
+    kernel_size,
+    p_dropout,
+    resblock, 
+    resblock_kernel_sizes, 
+    resblock_dilation_sizes, 
+    upsample_rates, 
+    upsample_initial_channel, 
+    upsample_kernel_sizes,
+    n_speakers=0,
+    gin_channels=0,
+    use_sdp=True,
+    **kwargs):
+
+    super().__init__()
+    self.n_vocab = n_vocab
+    self.spec_channels = spec_channels
+    self.inter_channels = inter_channels
+    self.hidden_channels = hidden_channels
+    self.filter_channels = filter_channels
+    self.n_heads = n_heads
+    self.n_layers = n_layers
+    self.kernel_size = kernel_size
+    self.p_dropout = p_dropout
+    self.resblock = resblock
+    self.resblock_kernel_sizes = resblock_kernel_sizes
+    self.resblock_dilation_sizes = resblock_dilation_sizes
+    self.upsample_rates = upsample_rates
+    self.upsample_initial_channel = upsample_initial_channel
+    self.upsample_kernel_sizes = upsample_kernel_sizes
+    self.segment_size = segment_size
+    self.n_speakers = n_speakers
+    self.gin_channels = gin_channels
+
+    self.use_sdp = use_sdp
+
+    self.enc_p = TextEncoder(n_vocab,
+        inter_channels,
+        hidden_channels,
+        filter_channels,
+        n_heads,
+        n_layers,
+        kernel_size,
+        p_dropout)
+    self.dec = Generator(inter_channels, resblock, resblock_kernel_sizes, resblock_dilation_sizes, upsample_rates, upsample_initial_channel, upsample_kernel_sizes, gin_channels=gin_channels)
+    self.enc_q = PosteriorEncoder(spec_channels, inter_channels, hidden_channels, 5, 1, 16, gin_channels=gin_channels)
+    self.flow = ResidualCouplingBlock(inter_channels, hidden_channels, 5, 1, 4, gin_channels=gin_channels)
+
+    if use_sdp:
+      self.dp = StochasticDurationPredictor(hidden_channels, 192, 3, 0.5, 4, gin_channels=gin_channels)
+    else:
+      self.dp = DurationPredictor(hidden_channels, 256, 3, 0.5, gin_channels=gin_channels)
+
+    if n_speakers > 1:
+      self.emb_g = nn.Embedding(n_speakers, gin_channels)
+
+  def forward(self, x, x_lengths, y, y_lengths, sid=None):
+
+    x, m_p, logs_p, x_mask = self.enc_p(x, x_lengths)
+    if self.n_speakers > 0:
+      g = self.emb_g(sid).unsqueeze(-1) # [b, h, 1]
+    else:
+      g = None
+
+    z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g)
+    z_p = self.flow(z, y_mask, g=g)
+
+    with torch.no_grad():
+      # negative cross-entropy
+      s_p_sq_r = torch.exp(-2 * logs_p) # [b, d, t]
+      neg_cent1 = torch.sum(-0.5 * math.log(2 * math.pi) - logs_p, [1], keepdim=True) # [b, 1, t_s]
+      neg_cent2 = torch.matmul(-0.5 * (z_p ** 2).transpose(1, 2), s_p_sq_r) # [b, t_t, d] x [b, d, t_s] = [b, t_t, t_s]
+      neg_cent3 = torch.matmul(z_p.transpose(1, 2), (m_p * s_p_sq_r)) # [b, t_t, d] x [b, d, t_s] = [b, t_t, t_s]
+      neg_cent4 = torch.sum(-0.5 * (m_p ** 2) * s_p_sq_r, [1], keepdim=True) # [b, 1, t_s]
+      neg_cent = neg_cent1 + neg_cent2 + neg_cent3 + neg_cent4
+
+      attn_mask = torch.unsqueeze(x_mask, 2) * torch.unsqueeze(y_mask, -1)
+      attn = monotonic_align.maximum_path(neg_cent, attn_mask.squeeze(1)).unsqueeze(1).detach()
+
+    w = attn.sum(2)
+    if self.use_sdp:
+      l_length = self.dp(x, x_mask, w, g=g)
+      l_length = l_length / torch.sum(x_mask)
+    else:
+      logw_ = torch.log(w + 1e-6) * x_mask
+      logw = self.dp(x, x_mask, g=g)
+      l_length = torch.sum((logw - logw_)**2, [1,2]) / torch.sum(x_mask) # for averaging 
+
+    # expand prior
+    m_p = torch.matmul(attn.squeeze(1), m_p.transpose(1, 2)).transpose(1, 2)
+    logs_p = torch.matmul(attn.squeeze(1), logs_p.transpose(1, 2)).transpose(1, 2)
+
+    z_slice, ids_slice = commons.rand_slice_segments(z, y_lengths, self.segment_size)
+    o = self.dec(z_slice, g=g)
+    return o, l_length, attn, ids_slice, x_mask, y_mask, (z, z_p, m_p, logs_p, m_q, logs_q)
+
+  def infer(self, x, x_lengths, sid=None, noise_scale=1, length_scale=1, noise_scale_w=1., max_len=None):
+    x, m_p, logs_p, x_mask = self.enc_p(x, x_lengths)
+    if self.n_speakers > 0:
+      g = self.emb_g(sid).unsqueeze(-1) # [b, h, 1]
+    else:
+      g = None
+
+    if self.use_sdp:
+      logw = self.dp(x, x_mask, g=g, reverse=True, noise_scale=noise_scale_w)
+    else:
+      logw = self.dp(x, x_mask, g=g)
+    w = torch.exp(logw) * x_mask * length_scale
+    w_ceil = torch.ceil(w)
+    y_lengths = torch.clamp_min(torch.sum(w_ceil, [1, 2]), 1).long()
+    y_mask = torch.unsqueeze(commons.sequence_mask(y_lengths, None), 1).to(x_mask.dtype)
+    attn_mask = torch.unsqueeze(x_mask, 2) * torch.unsqueeze(y_mask, -1)
+    attn = commons.generate_path(w_ceil, attn_mask)
+
+    m_p = torch.matmul(attn.squeeze(1), m_p.transpose(1, 2)).transpose(1, 2) # [b, t', t], [b, t, d] -> [b, d, t']
+    logs_p = torch.matmul(attn.squeeze(1), logs_p.transpose(1, 2)).transpose(1, 2) # [b, t', t], [b, t, d] -> [b, d, t']
+
+    z_p = m_p + torch.randn_like(m_p) * torch.exp(logs_p) * noise_scale
+    z = self.flow(z_p, y_mask, g=g, reverse=True)
+    o = self.dec((z * y_mask)[:,:,:max_len], g=g)
+    return o, attn, y_mask, (z, z_p, m_p, logs_p)
+
+  def voice_conversion(self, y, y_lengths, sid_src, sid_tgt):
+    assert self.n_speakers > 0, "n_speakers have to be larger than 0."
+    g_src = self.emb_g(sid_src).unsqueeze(-1)
+    g_tgt = self.emb_g(sid_tgt).unsqueeze(-1)
+    z, m_q, logs_q, y_mask = self.enc_q(y, y_lengths, g=g_src)
+    z_p = self.flow(z, y_mask, g=g_src)
+    z_hat = self.flow(z_p, y_mask, g=g_tgt, reverse=True)
+    o_hat = self.dec(z_hat * y_mask, g=g_tgt)
+    return o_hat, y_mask, (z, z_p, z_hat)
+

modules.py

@@ -0,0 +1,390 @@
+import copy
+import math
+import numpy as np
+import scipy
+import torch
+from torch import nn
+from torch.nn import functional as F
+
+from torch.nn import Conv1d, ConvTranspose1d, AvgPool1d, Conv2d
+from torch.nn.utils import weight_norm, remove_weight_norm
+
+import commons
+from commons import init_weights, get_padding
+from transforms import piecewise_rational_quadratic_transform
+
+
+LRELU_SLOPE = 0.1
+
+
+class LayerNorm(nn.Module):
+  def __init__(self, channels, eps=1e-5):
+    super().__init__()
+    self.channels = channels
+    self.eps = eps
+
+    self.gamma = nn.Parameter(torch.ones(channels))
+    self.beta = nn.Parameter(torch.zeros(channels))
+
+  def forward(self, x):
+    x = x.transpose(1, -1)
+    x = F.layer_norm(x, (self.channels,), self.gamma, self.beta, self.eps)
+    return x.transpose(1, -1)
+
+ 
+class ConvReluNorm(nn.Module):
+  def __init__(self, in_channels, hidden_channels, out_channels, kernel_size, n_layers, p_dropout):
+    super().__init__()
+    self.in_channels = in_channels
+    self.hidden_channels = hidden_channels
+    self.out_channels = out_channels
+    self.kernel_size = kernel_size
+    self.n_layers = n_layers
+    self.p_dropout = p_dropout
+    assert n_layers > 1, "Number of layers should be larger than 0."
+
+    self.conv_layers = nn.ModuleList()
+    self.norm_layers = nn.ModuleList()
+    self.conv_layers.append(nn.Conv1d(in_channels, hidden_channels, kernel_size, padding=kernel_size//2))
+    self.norm_layers.append(LayerNorm(hidden_channels))
+    self.relu_drop = nn.Sequential(
+        nn.ReLU(),
+        nn.Dropout(p_dropout))
+    for _ in range(n_layers-1):
+      self.conv_layers.append(nn.Conv1d(hidden_channels, hidden_channels, kernel_size, padding=kernel_size//2))
+      self.norm_layers.append(LayerNorm(hidden_channels))
+    self.proj = nn.Conv1d(hidden_channels, out_channels, 1)
+    self.proj.weight.data.zero_()
+    self.proj.bias.data.zero_()
+
+  def forward(self, x, x_mask):
+    x_org = x
+    for i in range(self.n_layers):
+      x = self.conv_layers[i](x * x_mask)
+      x = self.norm_layers[i](x)
+      x = self.relu_drop(x)
+    x = x_org + self.proj(x)
+    return x * x_mask
+
+
+class DDSConv(nn.Module):
+  """
+  Dialted and Depth-Separable Convolution
+  """
+  def __init__(self, channels, kernel_size, n_layers, p_dropout=0.):
+    super().__init__()
+    self.channels = channels
+    self.kernel_size = kernel_size
+    self.n_layers = n_layers
+    self.p_dropout = p_dropout
+
+    self.drop = nn.Dropout(p_dropout)
+    self.convs_sep = nn.ModuleList()
+    self.convs_1x1 = nn.ModuleList()
+    self.norms_1 = nn.ModuleList()
+    self.norms_2 = nn.ModuleList()
+    for i in range(n_layers):
+      dilation = kernel_size ** i
+      padding = (kernel_size * dilation - dilation) // 2
+      self.convs_sep.append(nn.Conv1d(channels, channels, kernel_size, 
+          groups=channels, dilation=dilation, padding=padding
+      ))
+      self.convs_1x1.append(nn.Conv1d(channels, channels, 1))
+      self.norms_1.append(LayerNorm(channels))
+      self.norms_2.append(LayerNorm(channels))
+
+  def forward(self, x, x_mask, g=None):
+    if g is not None:
+      x = x + g
+    for i in range(self.n_layers):
+      y = self.convs_sep[i](x * x_mask)
+      y = self.norms_1[i](y)
+      y = F.gelu(y)
+      y = self.convs_1x1[i](y)
+      y = self.norms_2[i](y)
+      y = F.gelu(y)
+      y = self.drop(y)
+      x = x + y
+    return x * x_mask
+
+
+class WN(torch.nn.Module):
+  def __init__(self, hidden_channels, kernel_size, dilation_rate, n_layers, gin_channels=0, p_dropout=0):
+    super(WN, self).__init__()
+    assert(kernel_size % 2 == 1)
+    self.hidden_channels =hidden_channels
+    self.kernel_size = kernel_size,
+    self.dilation_rate = dilation_rate
+    self.n_layers = n_layers
+    self.gin_channels = gin_channels
+    self.p_dropout = p_dropout
+
+    self.in_layers = torch.nn.ModuleList()
+    self.res_skip_layers = torch.nn.ModuleList()
+    self.drop = nn.Dropout(p_dropout)
+
+    if gin_channels != 0:
+      cond_layer = torch.nn.Conv1d(gin_channels, 2*hidden_channels*n_layers, 1)
+      self.cond_layer = torch.nn.utils.weight_norm(cond_layer, name='weight')
+
+    for i in range(n_layers):
+      dilation = dilation_rate ** i
+      padding = int((kernel_size * dilation - dilation) / 2)
+      in_layer = torch.nn.Conv1d(hidden_channels, 2*hidden_channels, kernel_size,
+                                 dilation=dilation, padding=padding)
+      in_layer = torch.nn.utils.weight_norm(in_layer, name='weight')
+      self.in_layers.append(in_layer)
+
+      # last one is not necessary
+      if i < n_layers - 1:
+        res_skip_channels = 2 * hidden_channels
+      else:
+        res_skip_channels = hidden_channels
+
+      res_skip_layer = torch.nn.Conv1d(hidden_channels, res_skip_channels, 1)
+      res_skip_layer = torch.nn.utils.weight_norm(res_skip_layer, name='weight')
+      self.res_skip_layers.append(res_skip_layer)
+
+  def forward(self, x, x_mask, g=None, **kwargs):
+    output = torch.zeros_like(x)
+    n_channels_tensor = torch.IntTensor([self.hidden_channels])
+
+    if g is not None:
+      g = self.cond_layer(g)
+
+    for i in range(self.n_layers):
+      x_in = self.in_layers[i](x)
+      if g is not None:
+        cond_offset = i * 2 * self.hidden_channels
+        g_l = g[:,cond_offset:cond_offset+2*self.hidden_channels,:]
+      else:
+        g_l = torch.zeros_like(x_in)
+
+      acts = commons.fused_add_tanh_sigmoid_multiply(
+          x_in,
+          g_l,
+          n_channels_tensor)
+      acts = self.drop(acts)
+
+      res_skip_acts = self.res_skip_layers[i](acts)
+      if i < self.n_layers - 1:
+        res_acts = res_skip_acts[:,:self.hidden_channels,:]
+        x = (x + res_acts) * x_mask
+        output = output + res_skip_acts[:,self.hidden_channels:,:]
+      else:
+        output = output + res_skip_acts
+    return output * x_mask
+
+  def remove_weight_norm(self):
+    if self.gin_channels != 0:
+      torch.nn.utils.remove_weight_norm(self.cond_layer)
+    for l in self.in_layers:
+      torch.nn.utils.remove_weight_norm(l)
+    for l in self.res_skip_layers:
+     torch.nn.utils.remove_weight_norm(l)
+
+
+class ResBlock1(torch.nn.Module):
+    def __init__(self, channels, kernel_size=3, dilation=(1, 3, 5)):
+        super(ResBlock1, self).__init__()
+        self.convs1 = nn.ModuleList([
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[0],
+                               padding=get_padding(kernel_size, dilation[0]))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[1],
+                               padding=get_padding(kernel_size, dilation[1]))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[2],
+                               padding=get_padding(kernel_size, dilation[2])))
+        ])
+        self.convs1.apply(init_weights)
+
+        self.convs2 = nn.ModuleList([
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=1,
+                               padding=get_padding(kernel_size, 1))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=1,
+                               padding=get_padding(kernel_size, 1))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=1,
+                               padding=get_padding(kernel_size, 1)))
+        ])
+        self.convs2.apply(init_weights)
+
+    def forward(self, x, x_mask=None):
+        for c1, c2 in zip(self.convs1, self.convs2):
+            xt = F.leaky_relu(x, LRELU_SLOPE)
+            if x_mask is not None:
+                xt = xt * x_mask
+            xt = c1(xt)
+            xt = F.leaky_relu(xt, LRELU_SLOPE)
+            if x_mask is not None:
+                xt = xt * x_mask
+            xt = c2(xt)
+            x = xt + x
+        if x_mask is not None:
+            x = x * x_mask
+        return x
+
+    def remove_weight_norm(self):
+        for l in self.convs1:
+            remove_weight_norm(l)
+        for l in self.convs2:
+            remove_weight_norm(l)
+
+
+class ResBlock2(torch.nn.Module):
+    def __init__(self, channels, kernel_size=3, dilation=(1, 3)):
+        super(ResBlock2, self).__init__()
+        self.convs = nn.ModuleList([
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[0],
+                               padding=get_padding(kernel_size, dilation[0]))),
+            weight_norm(Conv1d(channels, channels, kernel_size, 1, dilation=dilation[1],
+                               padding=get_padding(kernel_size, dilation[1])))
+        ])
+        self.convs.apply(init_weights)
+
+    def forward(self, x, x_mask=None):
+        for c in self.convs:
+            xt = F.leaky_relu(x, LRELU_SLOPE)
+            if x_mask is not None:
+                xt = xt * x_mask
+            xt = c(xt)
+            x = xt + x
+        if x_mask is not None:
+            x = x * x_mask
+        return x
+
+    def remove_weight_norm(self):
+        for l in self.convs:
+            remove_weight_norm(l)
+
+
+class Log(nn.Module):
+  def forward(self, x, x_mask, reverse=False, **kwargs):
+    if not reverse:
+      y = torch.log(torch.clamp_min(x, 1e-5)) * x_mask
+      logdet = torch.sum(-y, [1, 2])
+      return y, logdet
+    else:
+      x = torch.exp(x) * x_mask
+      return x
+    
+
+class Flip(nn.Module):
+  def forward(self, x, *args, reverse=False, **kwargs):
+    x = torch.flip(x, [1])
+    if not reverse:
+      logdet = torch.zeros(x.size(0)).to(dtype=x.dtype, device=x.device)
+      return x, logdet
+    else:
+      return x
+
+
+class ElementwiseAffine(nn.Module):
+  def __init__(self, channels):
+    super().__init__()
+    self.channels = channels
+    self.m = nn.Parameter(torch.zeros(channels,1))
+    self.logs = nn.Parameter(torch.zeros(channels,1))
+
+  def forward(self, x, x_mask, reverse=False, **kwargs):
+    if not reverse:
+      y = self.m + torch.exp(self.logs) * x
+      y = y * x_mask
+      logdet = torch.sum(self.logs * x_mask, [1,2])
+      return y, logdet
+    else:
+      x = (x - self.m) * torch.exp(-self.logs) * x_mask
+      return x
+
+
+class ResidualCouplingLayer(nn.Module):
+  def __init__(self,
+      channels,
+      hidden_channels,
+      kernel_size,
+      dilation_rate,
+      n_layers,
+      p_dropout=0,
+      gin_channels=0,
+      mean_only=False):
+    assert channels % 2 == 0, "channels should be divisible by 2"
+    super().__init__()
+    self.channels = channels
+    self.hidden_channels = hidden_channels
+    self.kernel_size = kernel_size
+    self.dilation_rate = dilation_rate
+    self.n_layers = n_layers
+    self.half_channels = channels // 2
+    self.mean_only = mean_only
+
+    self.pre = nn.Conv1d(self.half_channels, hidden_channels, 1)
+    self.enc = WN(hidden_channels, kernel_size, dilation_rate, n_layers, p_dropout=p_dropout, gin_channels=gin_channels)
+    self.post = nn.Conv1d(hidden_channels, self.half_channels * (2 - mean_only), 1)
+    self.post.weight.data.zero_()
+    self.post.bias.data.zero_()
+
+  def forward(self, x, x_mask, g=None, reverse=False):
+    x0, x1 = torch.split(x, [self.half_channels]*2, 1)
+    h = self.pre(x0) * x_mask
+    h = self.enc(h, x_mask, g=g)
+    stats = self.post(h) * x_mask
+    if not self.mean_only:
+      m, logs = torch.split(stats, [self.half_channels]*2, 1)
+    else:
+      m = stats
+      logs = torch.zeros_like(m)
+
+    if not reverse:
+      x1 = m + x1 * torch.exp(logs) * x_mask
+      x = torch.cat([x0, x1], 1)
+      logdet = torch.sum(logs, [1,2])
+      return x, logdet
+    else:
+      x1 = (x1 - m) * torch.exp(-logs) * x_mask
+      x = torch.cat([x0, x1], 1)
+      return x
+
+
+class ConvFlow(nn.Module):
+  def __init__(self, in_channels, filter_channels, kernel_size, n_layers, num_bins=10, tail_bound=5.0):
+    super().__init__()
+    self.in_channels = in_channels
+    self.filter_channels = filter_channels
+    self.kernel_size = kernel_size
+    self.n_layers = n_layers
+    self.num_bins = num_bins
+    self.tail_bound = tail_bound
+    self.half_channels = in_channels // 2
+
+    self.pre = nn.Conv1d(self.half_channels, filter_channels, 1)
+    self.convs = DDSConv(filter_channels, kernel_size, n_layers, p_dropout=0.)
+    self.proj = nn.Conv1d(filter_channels, self.half_channels * (num_bins * 3 - 1), 1)
+    self.proj.weight.data.zero_()
+    self.proj.bias.data.zero_()
+
+  def forward(self, x, x_mask, g=None, reverse=False):
+    x0, x1 = torch.split(x, [self.half_channels]*2, 1)
+    h = self.pre(x0)
+    h = self.convs(h, x_mask, g=g)
+    h = self.proj(h) * x_mask
+
+    b, c, t = x0.shape
+    h = h.reshape(b, c, -1, t).permute(0, 1, 3, 2) # [b, cx?, t] -> [b, c, t, ?]
+
+    unnormalized_widths = h[..., :self.num_bins] / math.sqrt(self.filter_channels)
+    unnormalized_heights = h[..., self.num_bins:2*self.num_bins] / math.sqrt(self.filter_channels)
+    unnormalized_derivatives = h[..., 2 * self.num_bins:]
+
+    x1, logabsdet = piecewise_rational_quadratic_transform(x1,
+        unnormalized_widths,
+        unnormalized_heights,
+        unnormalized_derivatives,
+        inverse=reverse,
+        tails='linear',
+        tail_bound=self.tail_bound
+    )
+
+    x = torch.cat([x0, x1], 1) * x_mask
+    logdet = torch.sum(logabsdet * x_mask, [1,2])
+    if not reverse:
+        return x, logdet
+    else:
+        return x

monotonic_align/__init__.py

@@ -0,0 +1,19 @@
+import numpy as np
+import torch
+from .monotonic_align.core import maximum_path_c
+
+
+def maximum_path(neg_cent, mask):
+  """ Cython optimized version.
+  neg_cent: [b, t_t, t_s]
+  mask: [b, t_t, t_s]
+  """
+  device = neg_cent.device
+  dtype = neg_cent.dtype
+  neg_cent = neg_cent.data.cpu().numpy().astype(np.float32)
+  path = np.zeros(neg_cent.shape, dtype=np.int32)
+
+  t_t_max = mask.sum(1)[:, 0].data.cpu().numpy().astype(np.int32)
+  t_s_max = mask.sum(2)[:, 0].data.cpu().numpy().astype(np.int32)
+  maximum_path_c(path, neg_cent, t_t_max, t_s_max)
+  return torch.from_numpy(path).to(device=device, dtype=dtype)

monotonic_align/core.pyx

@@ -0,0 +1,42 @@
+cimport cython
+from cython.parallel import prange
+
+
+@cython.boundscheck(False)
+@cython.wraparound(False)
+cdef void maximum_path_each(int[:,::1] path, float[:,::1] value, int t_y, int t_x, float max_neg_val=-1e9) nogil:
+  cdef int x
+  cdef int y
+  cdef float v_prev
+  cdef float v_cur
+  cdef float tmp
+  cdef int index = t_x - 1
+
+  for y in range(t_y):
+    for x in range(max(0, t_x + y - t_y), min(t_x, y + 1)):
+      if x == y:
+        v_cur = max_neg_val
+      else:
+        v_cur = value[y-1, x]
+      if x == 0:
+        if y == 0:
+          v_prev = 0.
+        else:
+          v_prev = max_neg_val
+      else:
+        v_prev = value[y-1, x-1]
+      value[y, x] += max(v_prev, v_cur)
+
+  for y in range(t_y - 1, -1, -1):
+    path[y, index] = 1
+    if index != 0 and (index == y or value[y-1, index] < value[y-1, index-1]):
+      index = index - 1
+
+
+@cython.boundscheck(False)
+@cython.wraparound(False)
+cpdef void maximum_path_c(int[:,:,::1] paths, float[:,:,::1] values, int[::1] t_ys, int[::1] t_xs) nogil:
+  cdef int b = paths.shape[0]
+  cdef int i
+  for i in prange(b, nogil=True):
+    maximum_path_each(paths[i], values[i], t_ys[i], t_xs[i])

monotonic_align/setup.py

@@ -0,0 +1,9 @@
+from distutils.core import setup
+from Cython.Build import cythonize
+import numpy
+
+setup(
+  name = 'monotonic_align',
+  ext_modules = cythonize("core.pyx"),
+  include_dirs=[numpy.get_include()]
+)

preprocess.py

@@ -0,0 +1,25 @@
+import argparse
+import text
+from utils import load_filepaths_and_text
+
+if __name__ == '__main__':
+  parser = argparse.ArgumentParser()
+  parser.add_argument("--out_extension", default="cleaned")
+  parser.add_argument("--text_index", default=2, type=int)
+  parser.add_argument("--filelists", nargs="+", default=["/content/vits/filelists/trainnn.txt"])
+  parser.add_argument("--text_cleaners", nargs="+", default=["japanese_cleaners"])
+
+  args = parser.parse_args()
+    
+
+  for filelist in args.filelists:
+    print("START:", filelist)
+    filepaths_and_text = load_filepaths_and_text(filelist)
+    for i in range(len(filepaths_and_text)):
+      original_text = filepaths_and_text[i][args.text_index]
+      cleaned_text = text._clean_text(original_text, args.text_cleaners)
+      filepaths_and_text[i][args.text_index] = cleaned_text
+
+    new_filelist = filelist + "." + args.out_extension
+    with open(new_filelist, "w", encoding="utf-8") as f:
+      f.writelines(["|".join(x) + "\n" for x in filepaths_and_text])

requirements.txt

@@ -0,0 +1,14 @@
+Cython==0.29.21
+librosa==0.8.0
+matplotlib==3.3.1
+numpy==1.21.6
+phonemizer==2.2.1
+scipy==1.5.2
+tensorboard==2.3.0
+torch==1.6.0
+torchvision==0.7.0
+Unidecode==1.1.1
+pyopenjtalk==0.2.0
+jamo==0.4.1
+pypinyin==0.44.0
+jieba==0.42.1

text/LICENSE

@@ -0,0 +1,19 @@
+Copyright (c) 2017 Keith Ito
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

text/__init__.py

@@ -0,0 +1,56 @@
+""" from https://github.com/keithito/tacotron """
+from text import cleaners
+from text.symbols import symbols
+
+
+# Mappings from symbol to numeric ID and vice versa:
+_symbol_to_id = {s: i for i, s in enumerate(symbols)}
+_id_to_symbol = {i: s for i, s in enumerate(symbols)}
+
+
+def text_to_sequence(text, cleaner_names):
+  '''Converts a string of text to a sequence of IDs corresponding to the symbols in the text.
+    Args:
+      text: string to convert to a sequence
+      cleaner_names: names of the cleaner functions to run the text through
+    Returns:
+      List of integers corresponding to the symbols in the text
+  '''
+  sequence = []
+
+  clean_text = _clean_text(text, cleaner_names)
+  for symbol in clean_text:
+    if symbol not in _symbol_to_id.keys():
+      continue
+    symbol_id = _symbol_to_id[symbol]
+    sequence += [symbol_id]
+  return sequence
+
+
+def cleaned_text_to_sequence(cleaned_text):
+  '''Converts a string of text to a sequence of IDs corresponding to the symbols in the text.
+    Args:
+      text: string to convert to a sequence
+    Returns:
+      List of integers corresponding to the symbols in the text
+  '''
+  sequence = [_symbol_to_id[symbol] for symbol in cleaned_text if symbol in _symbol_to_id.keys()]
+  return sequence
+
+
+def sequence_to_text(sequence):
+  '''Converts a sequence of IDs back to a string'''
+  result = ''
+  for symbol_id in sequence:
+    s = _id_to_symbol[symbol_id]
+    result += s
+  return result
+
+
+def _clean_text(text, cleaner_names):
+  for name in cleaner_names:
+    cleaner = getattr(cleaners, name)
+    if not cleaner:
+      raise Exception('Unknown cleaner: %s' % name)
+    text = cleaner(text)
+  return text

text/cleaners.py

@@ -0,0 +1,332 @@
+""" from https://github.com/keithito/tacotron """
+
+'''
+Cleaners are transformations that run over the input text at both training and eval time.
+
+Cleaners can be selected by passing a comma-delimited list of cleaner names as the "cleaners"
+hyperparameter. Some cleaners are English-specific. You'll typically want to use:
+  1. "english_cleaners" for English text
+  2. "transliteration_cleaners" for non-English text that can be transliterated to ASCII using
+     the Unidecode library (https://pypi.python.org/pypi/Unidecode)
+  3. "basic_cleaners" if you do not want to transliterate (in this case, you should also update
+     the symbols in symbols.py to match your data).
+'''
+
+import re
+from unidecode import unidecode
+import pyopenjtalk
+from jamo import h2j, j2hcj
+from pypinyin import lazy_pinyin,BOPOMOFO
+import jieba
+
+
+# This is a list of Korean classifiers preceded by pure Korean numerals.
+_korean_classifiers = '군데 권 개 그루 닢 대 두 마리 모 모금 뭇 발 발짝 방 번 벌 보루 살 수 술 시 쌈 움큼 정 짝 채 척 첩 축 켤레 톨 통'
+
+# Regular expression matching whitespace:
+_whitespace_re = re.compile(r'\s+')
+
+# Regular expression matching Japanese without punctuation marks:
+_japanese_characters = re.compile(r'[A-Za-z\d\u3005\u3040-\u30ff\u4e00-\u9fff\uff11-\uff19\uff21-\uff3a\uff41-\uff5a\uff66-\uff9d]')
+
+# Regular expression matching non-Japanese characters or punctuation marks:
+_japanese_marks = re.compile(r'[^A-Za-z\d\u3005\u3040-\u30ff\u4e00-\u9fff\uff11-\uff19\uff21-\uff3a\uff41-\uff5a\uff66-\uff9d]')
+
+# List of (regular expression, replacement) pairs for abbreviations:
+_abbreviations = [(re.compile('\\b%s\\.' % x[0], re.IGNORECASE), x[1]) for x in [
+  ('mrs', 'misess'),
+  ('mr', 'mister'),
+  ('dr', 'doctor'),
+  ('st', 'saint'),
+  ('co', 'company'),
+  ('jr', 'junior'),
+  ('maj', 'major'),
+  ('gen', 'general'),
+  ('drs', 'doctors'),
+  ('rev', 'reverend'),
+  ('lt', 'lieutenant'),
+  ('hon', 'honorable'),
+  ('sgt', 'sergeant'),
+  ('capt', 'captain'),
+  ('esq', 'esquire'),
+  ('ltd', 'limited'),
+  ('col', 'colonel'),
+  ('ft', 'fort'),
+]]
+
+# List of (hangul, hangul divided) pairs:
+_hangul_divided = [(re.compile('%s' % x[0]), x[1]) for x in [
+  ('ㄳ', 'ㄱㅅ'),
+  ('ㄵ', 'ㄴㅈ'),
+  ('ㄶ', 'ㄴㅎ'),
+  ('ㄺ', 'ㄹㄱ'),
+  ('ㄻ', 'ㄹㅁ'),
+  ('ㄼ', 'ㄹㅂ'),
+  ('ㄽ', 'ㄹㅅ'),
+  ('ㄾ', 'ㄹㅌ'),
+  ('ㄿ', 'ㄹㅍ'),
+  ('ㅀ', 'ㄹㅎ'),
+  ('ㅄ', 'ㅂㅅ'),
+  ('ㅘ', 'ㅗㅏ'),
+  ('ㅙ', 'ㅗㅐ'),
+  ('ㅚ', 'ㅗㅣ'),
+  ('ㅝ', 'ㅜㅓ'),
+  ('ㅞ', 'ㅜㅔ'),
+  ('ㅟ', 'ㅜㅣ'),
+  ('ㅢ', 'ㅡㅣ'),
+  ('ㅑ', 'ㅣㅏ'),
+  ('ㅒ', 'ㅣㅐ'),
+  ('ㅕ', 'ㅣㅓ'),
+  ('ㅖ', 'ㅣㅔ'),
+  ('ㅛ', 'ㅣㅗ'),
+  ('ㅠ', 'ㅣㅜ')
+]]
+
+# List of (Latin alphabet, hangul) pairs:
+_latin_to_hangul = [(re.compile('%s' % x[0], re.IGNORECASE), x[1]) for x in [
+  ('a', '에이'),
+  ('b', '비'),
+  ('c', '시'),
+  ('d', '디'),
+  ('e', '이'),
+  ('f', '에프'),
+  ('g', '지'),
+  ('h', '에이치'),
+  ('i', '아이'),
+  ('j', '제이'),
+  ('k', '케이'),
+  ('l', '엘'),
+  ('m', '엠'),
+  ('n', '엔'),
+  ('o', '오'),
+  ('p', '피'),
+  ('q', '큐'),
+  ('r', '아르'),
+  ('s', '에스'),
+  ('t', '티'),
+  ('u', '유'),
+  ('v', '브이'),
+  ('w', '더블유'),
+  ('x', '엑스'),
+  ('y', '와이'),
+  ('z', '제트')
+]]
+
+
+def expand_abbreviations(text):
+  for regex, replacement in _abbreviations:
+    text = re.sub(regex, replacement, text)
+  return text
+
+
+def lowercase(text):
+  return text.lower()
+
+
+def collapse_whitespace(text):
+  return re.sub(_whitespace_re, ' ', text)
+
+
+def convert_to_ascii(text):
+  return unidecode(text)
+
+
+def latin_to_hangul(text):
+  for regex, replacement in _latin_to_hangul:
+    text = re.sub(regex, replacement, text)
+  return text
+
+
+def divide_hangul(text):
+  for regex, replacement in _hangul_divided:
+    text = re.sub(regex, replacement, text)
+  return text
+
+
+def hangul_number(num, sino=True):
+  '''Reference https://github.com/Kyubyong/g2pK'''
+  num = re.sub(',', '', num)
+
+  if num == '0':
+      return '영'
+  if not sino and num == '20':
+      return '스무'
+
+  digits = '123456789'
+  names = '일이삼사오육칠팔구'
+  digit2name = {d: n for d, n in zip(digits, names)}
+  
+  modifiers = '한 두 세 네 다섯 여섯 일곱 여덟 아홉'
+  decimals = '열 스물 서른 마흔 쉰 예순 일흔 여든 아흔'
+  digit2mod = {d: mod for d, mod in zip(digits, modifiers.split())}
+  digit2dec = {d: dec for d, dec in zip(digits, decimals.split())}
+
+  spelledout = []
+  for i, digit in enumerate(num):
+    i = len(num) - i - 1
+    if sino:
+      if i == 0:
+        name = digit2name.get(digit, '')
+      elif i == 1:
+        name = digit2name.get(digit, '') + '십'
+        name = name.replace('일십', '십')
+    else:
+      if i == 0:
+        name = digit2mod.get(digit, '')
+      elif i == 1:
+        name = digit2dec.get(digit, '')
+    if digit == '0':
+      if i % 4 == 0:
+        last_three = spelledout[-min(3, len(spelledout)):]
+        if ''.join(last_three) == '':
+          spelledout.append('')
+          continue
+      else:
+        spelledout.append('')
+        continue
+    if i == 2:
+      name = digit2name.get(digit, '') + '백'
+      name = name.replace('일백', '백')
+    elif i == 3:
+      name = digit2name.get(digit, '') + '천'
+      name = name.replace('일천', '천')
+    elif i == 4:
+      name = digit2name.get(digit, '') + '만'
+      name = name.replace('일만', '만')
+    elif i == 5:
+      name = digit2name.get(digit, '') + '십'
+      name = name.replace('일십', '십')
+    elif i == 6:
+      name = digit2name.get(digit, '') + '백'
+      name = name.replace('일백', '백')
+    elif i == 7:
+      name = digit2name.get(digit, '') + '천'
+      name = name.replace('일천', '천')
+    elif i == 8:
+      name = digit2name.get(digit, '') + '억'
+    elif i == 9:
+      name = digit2name.get(digit, '') + '십'
+    elif i == 10:
+      name = digit2name.get(digit, '') + '백'
+    elif i == 11:
+      name = digit2name.get(digit, '') + '천'
+    elif i == 12:
+      name = digit2name.get(digit, '') + '조'
+    elif i == 13:
+      name = digit2name.get(digit, '') + '십'
+    elif i == 14:
+      name = digit2name.get(digit, '') + '백'
+    elif i == 15:
+      name = digit2name.get(digit, '') + '천'
+    spelledout.append(name)
+  return ''.join(elem for elem in spelledout)
+
+
+def number_to_hangul(text):
+  '''Reference https://github.com/Kyubyong/g2pK'''
+  tokens = set(re.findall(r'(\d[\d,]*)([\uac00-\ud71f]+)', text))
+  for token in tokens:
+    num, classifier = token
+    if classifier[:2] in _korean_classifiers or classifier[0] in _korean_classifiers:
+      spelledout = hangul_number(num, sino=False)
+    else:
+      spelledout = hangul_number(num, sino=True)
+    text = text.replace(f'{num}{classifier}', f'{spelledout}{classifier}')
+  # digit by digit for remaining digits
+  digits = '0123456789'
+  names = '영일이삼사오육칠팔구'
+  for d, n in zip(digits, names):
+    text = text.replace(d, n)
+  return text
+
+
+def basic_cleaners(text):
+  '''Basic pipeline that lowercases and collapses whitespace without transliteration.'''
+  text = lowercase(text)
+  text = collapse_whitespace(text)
+  return text
+
+
+def transliteration_cleaners(text):
+  '''Pipeline for non-English text that transliterates to ASCII.'''
+  text = convert_to_ascii(text)
+  text = lowercase(text)
+  text = collapse_whitespace(text)
+  return text
+
+
+def japanese_cleaners(text):
+  '''Pipeline for notating accent in Japanese text.
+  Reference https://r9y9.github.io/ttslearn/latest/notebooks/ch10_Recipe-Tacotron.html'''
+  sentences = re.split(_japanese_marks, text)
+  marks = re.findall(_japanese_marks, text)
+  text = ''
+  for i, sentence in enumerate(sentences):
+    if re.match(_japanese_characters, sentence):
+      if text!='':
+        text+=' '
+      labels = pyopenjtalk.extract_fullcontext(sentence)
+      for n, label in enumerate(labels):
+        phoneme = re.search(r'\-([^\+]*)\+', label).group(1)
+        if phoneme not in ['sil','pau']:
+          text += phoneme.replace('ch','ʧ').replace('sh','ʃ').replace('cl','Q')
+        else:
+          continue
+        n_moras = int(re.search(r'/F:(\d+)_', label).group(1))
+        a1 = int(re.search(r"/A:(\-?[0-9]+)\+", label).group(1))
+        a2 = int(re.search(r"\+(\d+)\+", label).group(1))
+        a3 = int(re.search(r"\+(\d+)/", label).group(1))
+        if re.search(r'\-([^\+]*)\+', labels[n + 1]).group(1) in ['sil','pau']:
+          a2_next=-1
+        else:
+          a2_next = int(re.search(r"\+(\d+)\+", labels[n + 1]).group(1))
+        # Accent phrase boundary
+        if a3 == 1 and a2_next == 1:
+          text += ' '
+        # Falling
+        elif a1 == 0 and a2_next == a2 + 1 and a2 != n_moras:
+          text += '↓'
+        # Rising
+        elif a2 == 1 and a2_next == 2:
+          text += '↑'
+    if i<len(marks):
+      text += unidecode(marks[i]).replace(' ','')
+  if re.match('[A-Za-z]',text[-1]):
+    text += '.'
+  return text
+
+
+def japanese_cleaners2(text):
+  return japanese_cleaners(text).replace('ts','ʦ').replace('...','…')
+
+
+def korean_cleaners(text):
+  '''Pipeline for Korean text'''
+  text = latin_to_hangul(text)
+  text = number_to_hangul(text)
+  text = j2hcj(h2j(text))
+  text = divide_hangul(text)
+  if re.match('[\u3131-\u3163]',text[-1]):
+    text += '.'
+  return text
+
+
+def chinese_cleaners(text):
+  '''Pipeline for Chinese text'''
+  text=text.replace('、','，').replace('；','，').replace('：','，')
+  words=jieba.lcut(text,cut_all=False)
+  text=''
+  for word in words:
+    bopomofos=lazy_pinyin(word,BOPOMOFO)
+    if not re.search('[\u4e00-\u9fff]',word):
+      text+=word
+      continue
+    for i in range(len(bopomofos)):
+      if re.match('[\u3105-\u3129]',bopomofos[i][-1]):
+        bopomofos[i]+='ˉ'
+    if text!='':
+      text+=' '
+    text+=''.join(bopomofos)
+  if re.match('[ˉˊˇˋ˙]',text[-1]):
+    text += '。'
+  return text

text/symbols.py

@@ -0,0 +1,33 @@
+'''
+Defines the set of symbols used in text input to the model.
+'''
+
+# japanese_cleaners
+_pad        = '_'
+_punctuation = ',.!?-'
+_letters = 'AEINOQUabdefghijkmnoprstuvwyzʃʧ↓↑ '
+
+
+# # japanese_cleaners2
+# _pad        = '_'
+# _punctuation = ',.!?-~…'
+# _letters = 'AEINOQUabdefghijkmnoprstuvwyzʃʧʦ↓↑ '
+
+
+'''# korean_cleaners
+_pad        = '_'
+_punctuation = ',.!?…~'
+_letters = 'ㄱㄴㄷㄹㅁㅂㅅㅇㅈㅊㅋㅌㅍㅎㄲㄸㅃㅆㅉㅏㅓㅗㅜㅡㅣㅐㅔ '
+'''
+
+'''# chinese_cleaners
+_pad        = '_'
+_punctuation = '，。！？—…'
+_letters = 'ㄅㄆㄇㄈㄉㄊㄋㄌㄍㄎㄏㄐㄑㄒㄓㄔㄕㄖㄗㄘㄙㄚㄛㄜㄝㄞㄟㄠㄡㄢㄣㄤㄥㄦㄧㄨㄩˉˊˇˋ˙ '
+'''
+
+# Export all symbols:
+symbols = [_pad] + list(_punctuation) + list(_letters)
+
+# Special symbol ids
+SPACE_ID = symbols.index(" ")

train.py

@@ -0,0 +1,295 @@
+import os
+import json
+import argparse
+import itertools
+import math
+import torch
+from torch import nn, optim
+from torch.nn import functional as F
+from torch.utils.data import DataLoader
+from torch.utils.tensorboard import SummaryWriter
+import torch.multiprocessing as mp
+import torch.distributed as dist
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.cuda.amp import autocast, GradScaler
+
+import librosa
+import logging
+
+logging.getLogger('numba').setLevel(logging.WARNING)
+
+import commons
+import utils
+from data_utils import (
+  TextAudioLoader,
+  TextAudioCollate,
+  DistributedBucketSampler
+)
+from models import (
+  SynthesizerTrn,
+  MultiPeriodDiscriminator,
+)
+from losses import (
+  generator_loss,
+  discriminator_loss,
+  feature_loss,
+  kl_loss
+)
+from mel_processing import mel_spectrogram_torch, spec_to_mel_torch
+from text.symbols import symbols
+
+
+torch.backends.cudnn.benchmark = True
+global_step = 0
+
+
+def main():
+  """Assume Single Node Multi GPUs Training Only"""
+  assert torch.cuda.is_available(), "CPU training is not allowed."
+
+  n_gpus = torch.cuda.device_count()
+  os.environ['MASTER_ADDR'] = 'localhost'
+  os.environ['MASTER_PORT'] = '80000'
+
+  hps = utils.get_hparams()
+  mp.spawn(run, nprocs=n_gpus, args=(n_gpus, hps,))
+
+
+def run(rank, n_gpus, hps):
+  global global_step
+  if rank == 0:
+    logger = utils.get_logger(hps.model_dir)
+    logger.info(hps)
+    utils.check_git_hash(hps.model_dir)
+    writer = SummaryWriter(log_dir=hps.model_dir)
+    writer_eval = SummaryWriter(log_dir=os.path.join(hps.model_dir, "eval"))
+
+  dist.init_process_group(backend='nccl', init_method='env://', world_size=n_gpus, rank=rank)
+  torch.manual_seed(hps.train.seed)
+  torch.cuda.set_device(rank)
+
+  train_dataset = TextAudioLoader(hps.data.training_files, hps.data)
+  train_sampler = DistributedBucketSampler(
+      train_dataset,
+      hps.train.batch_size,
+      [32,300,400,500,600,700,800,900,1000],
+      num_replicas=n_gpus,
+      rank=rank,
+      shuffle=True)
+  collate_fn = TextAudioCollate()
+  train_loader = DataLoader(train_dataset, num_workers=8, shuffle=False, pin_memory=True,
+      collate_fn=collate_fn, batch_sampler=train_sampler)
+  if rank == 0:
+    eval_dataset = TextAudioLoader(hps.data.validation_files, hps.data)
+    eval_loader = DataLoader(eval_dataset, num_workers=8, shuffle=False,
+        batch_size=hps.train.batch_size, pin_memory=True,
+        drop_last=False, collate_fn=collate_fn)
+
+  net_g = SynthesizerTrn(
+      len(symbols),
+      hps.data.filter_length // 2 + 1,
+      hps.train.segment_size // hps.data.hop_length,
+      **hps.model).cuda(rank)
+  net_d = MultiPeriodDiscriminator(hps.model.use_spectral_norm).cuda(rank)
+  optim_g = torch.optim.AdamW(
+      net_g.parameters(), 
+      hps.train.learning_rate, 
+      betas=hps.train.betas, 
+      eps=hps.train.eps)
+  optim_d = torch.optim.AdamW(
+      net_d.parameters(),
+      hps.train.learning_rate, 
+      betas=hps.train.betas, 
+      eps=hps.train.eps)
+  net_g = DDP(net_g, device_ids=[rank])
+  net_d = DDP(net_d, device_ids=[rank])
+
+  try:
+    _, _, _, epoch_str = utils.load_checkpoint(utils.latest_checkpoint_path(hps.model_dir, "G_*.pth"), net_g, optim_g)
+    _, _, _, epoch_str = utils.load_checkpoint(utils.latest_checkpoint_path(hps.model_dir, "D_*.pth"), net_d, optim_d)
+    global_step = (epoch_str - 1) * len(train_loader)
+  except:
+    epoch_str = 1
+    global_step = 0
+
+  scheduler_g = torch.optim.lr_scheduler.ExponentialLR(optim_g, gamma=hps.train.lr_decay, last_epoch=epoch_str-2)
+  scheduler_d = torch.optim.lr_scheduler.ExponentialLR(optim_d, gamma=hps.train.lr_decay, last_epoch=epoch_str-2)
+
+  scaler = GradScaler(enabled=hps.train.fp16_run)
+
+  for epoch in range(epoch_str, hps.train.epochs + 1):
+    if rank==0:
+      train_and_evaluate(rank, epoch, hps, [net_g, net_d], [optim_g, optim_d], [scheduler_g, scheduler_d], scaler, [train_loader, eval_loader], logger, [writer, writer_eval])
+    else:
+      train_and_evaluate(rank, epoch, hps, [net_g, net_d], [optim_g, optim_d], [scheduler_g, scheduler_d], scaler, [train_loader, None], None, None)
+    scheduler_g.step()
+    scheduler_d.step()
+
+
+def train_and_evaluate(rank, epoch, hps, nets, optims, schedulers, scaler, loaders, logger, writers):
+  net_g, net_d = nets
+  optim_g, optim_d = optims
+  scheduler_g, scheduler_d = schedulers
+  train_loader, eval_loader = loaders
+  if writers is not None:
+    writer, writer_eval = writers
+
+  train_loader.batch_sampler.set_epoch(epoch)
+  global global_step
+
+  net_g.train()
+  net_d.train()
+  for batch_idx, (x, x_lengths, spec, spec_lengths, y, y_lengths) in enumerate(train_loader):
+    x, x_lengths = x.cuda(rank, non_blocking=True), x_lengths.cuda(rank, non_blocking=True)
+    spec, spec_lengths = spec.cuda(rank, non_blocking=True), spec_lengths.cuda(rank, non_blocking=True)
+    y, y_lengths = y.cuda(rank, non_blocking=True), y_lengths.cuda(rank, non_blocking=True)
+
+    with autocast(enabled=hps.train.fp16_run):
+      y_hat, l_length, attn, ids_slice, x_mask, z_mask,\
+      (z, z_p, m_p, logs_p, m_q, logs_q) = net_g(x, x_lengths, spec, spec_lengths)
+
+      mel = spec_to_mel_torch(
+          spec, 
+          hps.data.filter_length, 
+          hps.data.n_mel_channels, 
+          hps.data.sampling_rate,
+          hps.data.mel_fmin, 
+          hps.data.mel_fmax)
+      y_mel = commons.slice_segments(mel, ids_slice, hps.train.segment_size // hps.data.hop_length)
+      y_hat_mel = mel_spectrogram_torch(
+          y_hat.squeeze(1), 
+          hps.data.filter_length, 
+          hps.data.n_mel_channels, 
+          hps.data.sampling_rate, 
+          hps.data.hop_length, 
+          hps.data.win_length, 
+          hps.data.mel_fmin, 
+          hps.data.mel_fmax
+      )
+
+      y = commons.slice_segments(y, ids_slice * hps.data.hop_length, hps.train.segment_size) # slice 
+
+      # Discriminator
+      y_d_hat_r, y_d_hat_g, _, _ = net_d(y, y_hat.detach())
+      with autocast(enabled=False):
+        loss_disc, losses_disc_r, losses_disc_g = discriminator_loss(y_d_hat_r, y_d_hat_g)
+        loss_disc_all = loss_disc
+    optim_d.zero_grad()
+    scaler.scale(loss_disc_all).backward()
+    scaler.unscale_(optim_d)
+    grad_norm_d = commons.clip_grad_value_(net_d.parameters(), None)
+    scaler.step(optim_d)
+
+    with autocast(enabled=hps.train.fp16_run):
+      # Generator
+      y_d_hat_r, y_d_hat_g, fmap_r, fmap_g = net_d(y, y_hat)
+      with autocast(enabled=False):
+        loss_dur = torch.sum(l_length.float())
+        loss_mel = F.l1_loss(y_mel, y_hat_mel) * hps.train.c_mel
+        loss_kl = kl_loss(z_p, logs_q, m_p, logs_p, z_mask) * hps.train.c_kl
+
+        loss_fm = feature_loss(fmap_r, fmap_g)
+        loss_gen, losses_gen = generator_loss(y_d_hat_g)
+        loss_gen_all = loss_gen + loss_fm + loss_mel + loss_dur + loss_kl
+    optim_g.zero_grad()
+    scaler.scale(loss_gen_all).backward()
+    scaler.unscale_(optim_g)
+    grad_norm_g = commons.clip_grad_value_(net_g.parameters(), None)
+    scaler.step(optim_g)
+    scaler.update()
+
+    if rank==0:
+      if global_step % hps.train.log_interval == 0:
+        lr = optim_g.param_groups[0]['lr']
+        losses = [loss_disc, loss_gen, loss_fm, loss_mel, loss_dur, loss_kl]
+        logger.info('Train Epoch: {} [{:.0f}%]'.format(
+          epoch,
+          100. * batch_idx / len(train_loader)))
+        logger.info([x.item() for x in losses] + [global_step, lr])
+        
+        scalar_dict = {"loss/g/total": loss_gen_all, "loss/d/total": loss_disc_all, "learning_rate": lr, "grad_norm_d": grad_norm_d, "grad_norm_g": grad_norm_g}
+        scalar_dict.update({"loss/g/fm": loss_fm, "loss/g/mel": loss_mel, "loss/g/dur": loss_dur, "loss/g/kl": loss_kl})
+
+        scalar_dict.update({"loss/g/{}".format(i): v for i, v in enumerate(losses_gen)})
+        scalar_dict.update({"loss/d_r/{}".format(i): v for i, v in enumerate(losses_disc_r)})
+        scalar_dict.update({"loss/d_g/{}".format(i): v for i, v in enumerate(losses_disc_g)})
+        image_dict = { 
+            "slice/mel_org": utils.plot_spectrogram_to_numpy(y_mel[0].data.cpu().numpy()),
+            "slice/mel_gen": utils.plot_spectrogram_to_numpy(y_hat_mel[0].data.cpu().numpy()), 
+            "all/mel": utils.plot_spectrogram_to_numpy(mel[0].data.cpu().numpy()),
+            "all/attn": utils.plot_alignment_to_numpy(attn[0,0].data.cpu().numpy())
+        }
+        utils.summarize(
+          writer=writer,
+          global_step=global_step, 
+          images=image_dict,
+          scalars=scalar_dict)
+
+      if global_step % hps.train.eval_interval == 0:
+        evaluate(hps, net_g, eval_loader, writer_eval)
+        utils.save_checkpoint(net_g, optim_g, hps.train.learning_rate, epoch, os.path.join(hps.model_dir, "G_{}.pth".format(global_step)))
+        utils.save_checkpoint(net_d, optim_d, hps.train.learning_rate, epoch, os.path.join(hps.model_dir, "D_{}.pth".format(global_step)))
+    global_step += 1
+  
+  if rank == 0:
+    logger.info('====> Epoch: {}'.format(epoch))
+
+ 
+def evaluate(hps, generator, eval_loader, writer_eval):
+    generator.eval()
+    with torch.no_grad():
+      for batch_idx, (x, x_lengths, spec, spec_lengths, y, y_lengths) in enumerate(eval_loader):
+        x, x_lengths = x.cuda(0), x_lengths.cuda(0)
+        spec, spec_lengths = spec.cuda(0), spec_lengths.cuda(0)
+        y, y_lengths = y.cuda(0), y_lengths.cuda(0)
+
+        # remove else
+        x = x[:1]
+        x_lengths = x_lengths[:1]
+        spec = spec[:1]
+        spec_lengths = spec_lengths[:1]
+        y = y[:1]
+        y_lengths = y_lengths[:1]
+        break
+      y_hat, attn, mask, *_ = generator.module.infer(x, x_lengths, max_len=1000)
+      y_hat_lengths = mask.sum([1,2]).long() * hps.data.hop_length
+
+      mel = spec_to_mel_torch(
+        spec, 
+        hps.data.filter_length, 
+        hps.data.n_mel_channels, 
+        hps.data.sampling_rate,
+        hps.data.mel_fmin, 
+        hps.data.mel_fmax)
+      y_hat_mel = mel_spectrogram_torch(
+        y_hat.squeeze(1).float(),
+        hps.data.filter_length,
+        hps.data.n_mel_channels,
+        hps.data.sampling_rate,
+        hps.data.hop_length,
+        hps.data.win_length,
+        hps.data.mel_fmin,
+        hps.data.mel_fmax
+      )
+    image_dict = {
+      "gen/mel": utils.plot_spectrogram_to_numpy(y_hat_mel[0].cpu().numpy())
+    }
+    audio_dict = {
+      "gen/audio": y_hat[0,:,:y_hat_lengths[0]]
+    }
+    if global_step == 0:
+      image_dict.update({"gt/mel": utils.plot_spectrogram_to_numpy(mel[0].cpu().numpy())})
+      audio_dict.update({"gt/audio": y[0,:,:y_lengths[0]]})
+
+    utils.summarize(
+      writer=writer_eval,
+      global_step=global_step, 
+      images=image_dict,
+      audios=audio_dict,
+      audio_sampling_rate=hps.data.sampling_rate
+    )
+    generator.train()
+
+                           
+if __name__ == "__main__":
+  main()

train_ms.py

@@ -0,0 +1,299 @@
+import os
+import json
+import argparse
+import itertools
+import math
+import torch
+from torch import nn, optim
+from torch.nn import functional as F
+from torch.utils.data import DataLoader
+from torch.utils.tensorboard import SummaryWriter
+import torch.multiprocessing as mp
+import torch.distributed as dist
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.cuda.amp import autocast, GradScaler
+
+import librosa
+import logging
+
+logging.getLogger('numba').setLevel(logging.WARNING)
+
+import commons
+import utils
+from data_utils import (
+  TextAudioSpeakerLoader,
+  TextAudioSpeakerCollate,
+  DistributedBucketSampler
+)
+from models import (
+  SynthesizerTrn,
+  MultiPeriodDiscriminator,
+)
+from losses import (
+  generator_loss,
+  discriminator_loss,
+  feature_loss,
+  kl_loss
+)
+from mel_processing import mel_spectrogram_torch, spec_to_mel_torch
+from text.symbols import symbols
+
+
+torch.backends.cudnn.benchmark = True
+global_step = 0
+
+
+def main():
+  """Assume Single Node Multi GPUs Training Only"""
+  assert torch.cuda.is_available(), "CPU training is not allowed."
+
+  n_gpus = torch.cuda.device_count()
+  os.environ['MASTER_ADDR'] = 'localhost'
+  os.environ['MASTER_PORT'] = '80000'
+
+  hps = utils.get_hparams()
+  mp.spawn(run, nprocs=n_gpus, args=(n_gpus, hps,))
+
+
+def run(rank, n_gpus, hps):
+  global global_step
+  if rank == 0:
+    logger = utils.get_logger(hps.model_dir)
+    logger.info(hps)
+    utils.check_git_hash(hps.model_dir)
+    writer = SummaryWriter(log_dir=hps.model_dir)
+    writer_eval = SummaryWriter(log_dir=os.path.join(hps.model_dir, "eval"))
+
+  dist.init_process_group(backend='nccl', init_method='env://', world_size=n_gpus, rank=rank)
+  torch.manual_seed(hps.train.seed)
+  torch.cuda.set_device(rank)
+
+  train_dataset = TextAudioSpeakerLoader(hps.data.training_files, hps.data)
+  train_sampler = DistributedBucketSampler(
+      train_dataset,
+      hps.train.batch_size,
+      [32,300,400,500,600,700,800,900,1000],
+      num_replicas=n_gpus,
+      rank=rank,
+      shuffle=True)
+  collate_fn = TextAudioSpeakerCollate()
+  train_loader = DataLoader(train_dataset, num_workers=8, shuffle=False, pin_memory=True,
+      collate_fn=collate_fn, batch_sampler=train_sampler)
+  if rank == 0:
+    eval_dataset = TextAudioSpeakerLoader(hps.data.validation_files, hps.data)
+    eval_loader = DataLoader(eval_dataset, num_workers=8, shuffle=False,
+        batch_size=hps.train.batch_size, pin_memory=True,
+        drop_last=False, collate_fn=collate_fn)
+
+  net_g = SynthesizerTrn(
+      len(symbols),
+      hps.data.filter_length // 2 + 1,
+      hps.train.segment_size // hps.data.hop_length,
+      n_speakers=hps.data.n_speakers,
+      **hps.model).cuda(rank)
+  net_d = MultiPeriodDiscriminator(hps.model.use_spectral_norm).cuda(rank)
+  optim_g = torch.optim.AdamW(
+      net_g.parameters(), 
+      hps.train.learning_rate, 
+      betas=hps.train.betas, 
+      eps=hps.train.eps)
+  optim_d = torch.optim.AdamW(
+      net_d.parameters(),
+      hps.train.learning_rate, 
+      betas=hps.train.betas, 
+      eps=hps.train.eps)
+  net_g = DDP(net_g, device_ids=[rank])
+  net_d = DDP(net_d, device_ids=[rank])
+
+  try:
+    _, _, _, epoch_str = utils.load_checkpoint(utils.latest_checkpoint_path(hps.model_dir, "G_*.pth"), net_g, optim_g)
+    _, _, _, epoch_str = utils.load_checkpoint(utils.latest_checkpoint_path(hps.model_dir, "D_*.pth"), net_d, optim_d)
+    global_step = (epoch_str - 1) * len(train_loader)
+  except:
+    epoch_str = 1
+    global_step = 0
+
+  scheduler_g = torch.optim.lr_scheduler.ExponentialLR(optim_g, gamma=hps.train.lr_decay, last_epoch=epoch_str-2)
+  scheduler_d = torch.optim.lr_scheduler.ExponentialLR(optim_d, gamma=hps.train.lr_decay, last_epoch=epoch_str-2)
+
+  scaler = GradScaler(enabled=hps.train.fp16_run)
+
+  for epoch in range(epoch_str, hps.train.epochs + 1):
+    if rank==0:
+      train_and_evaluate(rank, epoch, hps, [net_g, net_d], [optim_g, optim_d], [scheduler_g, scheduler_d], scaler, [train_loader, eval_loader], logger, [writer, writer_eval])
+    else:
+      train_and_evaluate(rank, epoch, hps, [net_g, net_d], [optim_g, optim_d], [scheduler_g, scheduler_d], scaler, [train_loader, None], None, None)
+    scheduler_g.step()
+    scheduler_d.step()
+
+
+def train_and_evaluate(rank, epoch, hps, nets, optims, schedulers, scaler, loaders, logger, writers):
+  net_g, net_d = nets
+  optim_g, optim_d = optims
+  scheduler_g, scheduler_d = schedulers
+  train_loader, eval_loader = loaders
+  if writers is not None:
+    writer, writer_eval = writers
+
+  train_loader.batch_sampler.set_epoch(epoch)
+  global global_step
+
+  net_g.train()
+  net_d.train()
+  for batch_idx, (x, x_lengths, spec, spec_lengths, y, y_lengths, speakers) in enumerate(train_loader):
+    x, x_lengths = x.cuda(rank, non_blocking=True), x_lengths.cuda(rank, non_blocking=True)
+    spec, spec_lengths = spec.cuda(rank, non_blocking=True), spec_lengths.cuda(rank, non_blocking=True)
+    y, y_lengths = y.cuda(rank, non_blocking=True), y_lengths.cuda(rank, non_blocking=True)
+    speakers = speakers.cuda(rank, non_blocking=True)
+
+    with autocast(enabled=hps.train.fp16_run):
+      y_hat, l_length, attn, ids_slice, x_mask, z_mask,\
+      (z, z_p, m_p, logs_p, m_q, logs_q) = net_g(x, x_lengths, spec, spec_lengths, speakers)
+
+      mel = spec_to_mel_torch(
+          spec, 
+          hps.data.filter_length, 
+          hps.data.n_mel_channels, 
+          hps.data.sampling_rate,
+          hps.data.mel_fmin, 
+          hps.data.mel_fmax)
+      y_mel = commons.slice_segments(mel, ids_slice, hps.train.segment_size // hps.data.hop_length)
+      y_hat_mel = mel_spectrogram_torch(
+          y_hat.squeeze(1), 
+          hps.data.filter_length, 
+          hps.data.n_mel_channels, 
+          hps.data.sampling_rate, 
+          hps.data.hop_length, 
+          hps.data.win_length, 
+          hps.data.mel_fmin, 
+          hps.data.mel_fmax
+      )
+
+      y = commons.slice_segments(y, ids_slice * hps.data.hop_length, hps.train.segment_size) # slice 
+
+      # Discriminator
+      y_d_hat_r, y_d_hat_g, _, _ = net_d(y, y_hat.detach())
+      with autocast(enabled=False):
+        loss_disc, losses_disc_r, losses_disc_g = discriminator_loss(y_d_hat_r, y_d_hat_g)
+        loss_disc_all = loss_disc
+    optim_d.zero_grad()
+    scaler.scale(loss_disc_all).backward()
+    scaler.unscale_(optim_d)
+    grad_norm_d = commons.clip_grad_value_(net_d.parameters(), None)
+    scaler.step(optim_d)
+
+    with autocast(enabled=hps.train.fp16_run):
+      # Generator
+      y_d_hat_r, y_d_hat_g, fmap_r, fmap_g = net_d(y, y_hat)
+      with autocast(enabled=False):
+        loss_dur = torch.sum(l_length.float())
+        loss_mel = F.l1_loss(y_mel, y_hat_mel) * hps.train.c_mel
+        loss_kl = kl_loss(z_p, logs_q, m_p, logs_p, z_mask) * hps.train.c_kl
+
+        loss_fm = feature_loss(fmap_r, fmap_g)
+        loss_gen, losses_gen = generator_loss(y_d_hat_g)
+        loss_gen_all = loss_gen + loss_fm + loss_mel + loss_dur + loss_kl
+    optim_g.zero_grad()
+    scaler.scale(loss_gen_all).backward()
+    scaler.unscale_(optim_g)
+    grad_norm_g = commons.clip_grad_value_(net_g.parameters(), None)
+    scaler.step(optim_g)
+    scaler.update()
+
+    if rank==0:
+      if global_step % hps.train.log_interval == 0:
+        lr = optim_g.param_groups[0]['lr']
+        losses = [loss_disc, loss_gen, loss_fm, loss_mel, loss_dur, loss_kl]
+        logger.info('Train Epoch: {} [{:.0f}%]'.format(
+          epoch,
+          100. * batch_idx / len(train_loader)))
+        logger.info([x.item() for x in losses] + [global_step, lr])
+        
+        scalar_dict = {"loss/g/total": loss_gen_all, "loss/d/total": loss_disc_all, "learning_rate": lr, "grad_norm_d": grad_norm_d, "grad_norm_g": grad_norm_g}
+        scalar_dict.update({"loss/g/fm": loss_fm, "loss/g/mel": loss_mel, "loss/g/dur": loss_dur, "loss/g/kl": loss_kl})
+
+        scalar_dict.update({"loss/g/{}".format(i): v for i, v in enumerate(losses_gen)})
+        scalar_dict.update({"loss/d_r/{}".format(i): v for i, v in enumerate(losses_disc_r)})
+        scalar_dict.update({"loss/d_g/{}".format(i): v for i, v in enumerate(losses_disc_g)})
+        image_dict = { 
+            "slice/mel_org": utils.plot_spectrogram_to_numpy(y_mel[0].data.cpu().numpy()),
+            "slice/mel_gen": utils.plot_spectrogram_to_numpy(y_hat_mel[0].data.cpu().numpy()), 
+            "all/mel": utils.plot_spectrogram_to_numpy(mel[0].data.cpu().numpy()),
+            "all/attn": utils.plot_alignment_to_numpy(attn[0,0].data.cpu().numpy())
+        }
+        utils.summarize(
+          writer=writer,
+          global_step=global_step, 
+          images=image_dict,
+          scalars=scalar_dict)
+
+      if global_step % hps.train.eval_interval == 0:
+        evaluate(hps, net_g, eval_loader, writer_eval)
+        utils.save_checkpoint(net_g, optim_g, hps.train.learning_rate, epoch, os.path.join(hps.model_dir, "G_{}.pth".format(global_step)))
+        utils.save_checkpoint(net_d, optim_d, hps.train.learning_rate, epoch, os.path.join(hps.model_dir, "D_{}.pth".format(global_step)))
+    global_step += 1
+  
+  if rank == 0:
+    logger.info('====> Epoch: {}'.format(epoch))
+
+ 
+def evaluate(hps, generator, eval_loader, writer_eval):
+    generator.eval()
+    with torch.no_grad():
+      for batch_idx, (x, x_lengths, spec, spec_lengths, y, y_lengths, speakers) in enumerate(eval_loader):
+        x, x_lengths = x.cuda(0), x_lengths.cuda(0)
+        spec, spec_lengths = spec.cuda(0), spec_lengths.cuda(0)
+        y, y_lengths = y.cuda(0), y_lengths.cuda(0)
+        speakers = speakers.cuda(0)
+
+        # remove else
+        x = x[:1]
+        x_lengths = x_lengths[:1]
+        spec = spec[:1]
+        spec_lengths = spec_lengths[:1]
+        y = y[:1]
+        y_lengths = y_lengths[:1]
+        speakers = speakers[:1]
+        break
+      y_hat, attn, mask, *_ = generator.module.infer(x, x_lengths, speakers, max_len=1000)
+      y_hat_lengths = mask.sum([1,2]).long() * hps.data.hop_length
+
+      mel = spec_to_mel_torch(
+        spec, 
+        hps.data.filter_length, 
+        hps.data.n_mel_channels, 
+        hps.data.sampling_rate,
+        hps.data.mel_fmin, 
+        hps.data.mel_fmax)
+      y_hat_mel = mel_spectrogram_torch(
+        y_hat.squeeze(1).float(),
+        hps.data.filter_length,
+        hps.data.n_mel_channels,
+        hps.data.sampling_rate,
+        hps.data.hop_length,
+        hps.data.win_length,
+        hps.data.mel_fmin,
+        hps.data.mel_fmax
+      )
+    image_dict = {
+      "gen/mel": utils.plot_spectrogram_to_numpy(y_hat_mel[0].cpu().numpy())
+    }
+    audio_dict = {
+      "gen/audio": y_hat[0,:,:y_hat_lengths[0]]
+    }
+    if global_step == 0:
+      image_dict.update({"gt/mel": utils.plot_spectrogram_to_numpy(mel[0].cpu().numpy())})
+      audio_dict.update({"gt/audio": y[0,:,:y_lengths[0]]})
+
+    utils.summarize(
+      writer=writer_eval,
+      global_step=global_step, 
+      images=image_dict,
+      audios=audio_dict,
+      audio_sampling_rate=hps.data.sampling_rate
+    )
+    generator.train()
+
+                           
+if __name__ == "__main__":
+  main()

transforms.py

@@ -0,0 +1,193 @@
+import torch
+from torch.nn import functional as F
+
+import numpy as np
+
+
+DEFAULT_MIN_BIN_WIDTH = 1e-3
+DEFAULT_MIN_BIN_HEIGHT = 1e-3
+DEFAULT_MIN_DERIVATIVE = 1e-3
+
+
+def piecewise_rational_quadratic_transform(inputs, 
+                                           unnormalized_widths,
+                                           unnormalized_heights,
+                                           unnormalized_derivatives,
+                                           inverse=False,
+                                           tails=None, 
+                                           tail_bound=1.,
+                                           min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+                                           min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+                                           min_derivative=DEFAULT_MIN_DERIVATIVE):
+
+    if tails is None:
+        spline_fn = rational_quadratic_spline
+        spline_kwargs = {}
+    else:
+        spline_fn = unconstrained_rational_quadratic_spline
+        spline_kwargs = {
+            'tails': tails,
+            'tail_bound': tail_bound
+        }
+
+    outputs, logabsdet = spline_fn(
+            inputs=inputs,
+            unnormalized_widths=unnormalized_widths,
+            unnormalized_heights=unnormalized_heights,
+            unnormalized_derivatives=unnormalized_derivatives,
+            inverse=inverse,
+            min_bin_width=min_bin_width,
+            min_bin_height=min_bin_height,
+            min_derivative=min_derivative,
+            **spline_kwargs
+    )
+    return outputs, logabsdet
+
+
+def searchsorted(bin_locations, inputs, eps=1e-6):
+    bin_locations[..., -1] += eps
+    return torch.sum(
+        inputs[..., None] >= bin_locations,
+        dim=-1
+    ) - 1
+
+
+def unconstrained_rational_quadratic_spline(inputs,
+                                            unnormalized_widths,
+                                            unnormalized_heights,
+                                            unnormalized_derivatives,
+                                            inverse=False,
+                                            tails='linear',
+                                            tail_bound=1.,
+                                            min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+                                            min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+                                            min_derivative=DEFAULT_MIN_DERIVATIVE):
+    inside_interval_mask = (inputs >= -tail_bound) & (inputs <= tail_bound)
+    outside_interval_mask = ~inside_interval_mask
+
+    outputs = torch.zeros_like(inputs)
+    logabsdet = torch.zeros_like(inputs)
+
+    if tails == 'linear':
+        unnormalized_derivatives = F.pad(unnormalized_derivatives, pad=(1, 1))
+        constant = np.log(np.exp(1 - min_derivative) - 1)
+        unnormalized_derivatives[..., 0] = constant
+        unnormalized_derivatives[..., -1] = constant
+
+        outputs[outside_interval_mask] = inputs[outside_interval_mask]
+        logabsdet[outside_interval_mask] = 0
+    else:
+        raise RuntimeError('{} tails are not implemented.'.format(tails))
+
+    outputs[inside_interval_mask], logabsdet[inside_interval_mask] = rational_quadratic_spline(
+        inputs=inputs[inside_interval_mask],
+        unnormalized_widths=unnormalized_widths[inside_interval_mask, :],
+        unnormalized_heights=unnormalized_heights[inside_interval_mask, :],
+        unnormalized_derivatives=unnormalized_derivatives[inside_interval_mask, :],
+        inverse=inverse,
+        left=-tail_bound, right=tail_bound, bottom=-tail_bound, top=tail_bound,
+        min_bin_width=min_bin_width,
+        min_bin_height=min_bin_height,
+        min_derivative=min_derivative
+    )
+
+    return outputs, logabsdet
+
+def rational_quadratic_spline(inputs,
+                              unnormalized_widths,
+                              unnormalized_heights,
+                              unnormalized_derivatives,
+                              inverse=False,
+                              left=0., right=1., bottom=0., top=1.,
+                              min_bin_width=DEFAULT_MIN_BIN_WIDTH,
+                              min_bin_height=DEFAULT_MIN_BIN_HEIGHT,
+                              min_derivative=DEFAULT_MIN_DERIVATIVE):
+    if torch.min(inputs) < left or torch.max(inputs) > right:
+        raise ValueError('Input to a transform is not within its domain')
+
+    num_bins = unnormalized_widths.shape[-1]
+
+    if min_bin_width * num_bins > 1.0:
+        raise ValueError('Minimal bin width too large for the number of bins')
+    if min_bin_height * num_bins > 1.0:
+        raise ValueError('Minimal bin height too large for the number of bins')
+
+    widths = F.softmax(unnormalized_widths, dim=-1)
+    widths = min_bin_width + (1 - min_bin_width * num_bins) * widths
+    cumwidths = torch.cumsum(widths, dim=-1)
+    cumwidths = F.pad(cumwidths, pad=(1, 0), mode='constant', value=0.0)
+    cumwidths = (right - left) * cumwidths + left
+    cumwidths[..., 0] = left
+    cumwidths[..., -1] = right
+    widths = cumwidths[..., 1:] - cumwidths[..., :-1]
+
+    derivatives = min_derivative + F.softplus(unnormalized_derivatives)
+
+    heights = F.softmax(unnormalized_heights, dim=-1)
+    heights = min_bin_height + (1 - min_bin_height * num_bins) * heights
+    cumheights = torch.cumsum(heights, dim=-1)
+    cumheights = F.pad(cumheights, pad=(1, 0), mode='constant', value=0.0)
+    cumheights = (top - bottom) * cumheights + bottom
+    cumheights[..., 0] = bottom
+    cumheights[..., -1] = top
+    heights = cumheights[..., 1:] - cumheights[..., :-1]
+
+    if inverse:
+        bin_idx = searchsorted(cumheights, inputs)[..., None]
+    else:
+        bin_idx = searchsorted(cumwidths, inputs)[..., None]
+
+    input_cumwidths = cumwidths.gather(-1, bin_idx)[..., 0]
+    input_bin_widths = widths.gather(-1, bin_idx)[..., 0]
+
+    input_cumheights = cumheights.gather(-1, bin_idx)[..., 0]
+    delta = heights / widths
+    input_delta = delta.gather(-1, bin_idx)[..., 0]
+
+    input_derivatives = derivatives.gather(-1, bin_idx)[..., 0]
+    input_derivatives_plus_one = derivatives[..., 1:].gather(-1, bin_idx)[..., 0]
+
+    input_heights = heights.gather(-1, bin_idx)[..., 0]
+
+    if inverse:
+        a = (((inputs - input_cumheights) * (input_derivatives
+                                             + input_derivatives_plus_one
+                                             - 2 * input_delta)
+              + input_heights * (input_delta - input_derivatives)))
+        b = (input_heights * input_derivatives
+             - (inputs - input_cumheights) * (input_derivatives
+                                              + input_derivatives_plus_one
+                                              - 2 * input_delta))
+        c = - input_delta * (inputs - input_cumheights)
+
+        discriminant = b.pow(2) - 4 * a * c
+        assert (discriminant >= 0).all()
+
+        root = (2 * c) / (-b - torch.sqrt(discriminant))
+        outputs = root * input_bin_widths + input_cumwidths
+
+        theta_one_minus_theta = root * (1 - root)
+        denominator = input_delta + ((input_derivatives + input_derivatives_plus_one - 2 * input_delta)
+                                     * theta_one_minus_theta)
+        derivative_numerator = input_delta.pow(2) * (input_derivatives_plus_one * root.pow(2)
+                                                     + 2 * input_delta * theta_one_minus_theta
+                                                     + input_derivatives * (1 - root).pow(2))
+        logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator)
+
+        return outputs, -logabsdet
+    else:
+        theta = (inputs - input_cumwidths) / input_bin_widths
+        theta_one_minus_theta = theta * (1 - theta)
+
+        numerator = input_heights * (input_delta * theta.pow(2)
+                                     + input_derivatives * theta_one_minus_theta)
+        denominator = input_delta + ((input_derivatives + input_derivatives_plus_one - 2 * input_delta)
+                                     * theta_one_minus_theta)
+        outputs = input_cumheights + numerator / denominator
+
+        derivative_numerator = input_delta.pow(2) * (input_derivatives_plus_one * theta.pow(2)
+                                                     + 2 * input_delta * theta_one_minus_theta
+                                                     + input_derivatives * (1 - theta).pow(2))
+        logabsdet = torch.log(derivative_numerator) - 2 * torch.log(denominator)
+
+        return outputs, logabsdet

utils.py

@@ -0,0 +1,258 @@
+import os
+import glob
+import sys
+import argparse
+import logging
+import json
+import subprocess
+import numpy as np
+from scipy.io.wavfile import read
+import torch
+
+MATPLOTLIB_FLAG = False
+
+logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)
+logger = logging
+
+
+def load_checkpoint(checkpoint_path, model, optimizer=None):
+  assert os.path.isfile(checkpoint_path)
+  checkpoint_dict = torch.load(checkpoint_path, map_location='cpu')
+  iteration = checkpoint_dict['iteration']
+  learning_rate = checkpoint_dict['learning_rate']
+  if optimizer is not None:
+    optimizer.load_state_dict(checkpoint_dict['optimizer'])
+  saved_state_dict = checkpoint_dict['model']
+  if hasattr(model, 'module'):
+    state_dict = model.module.state_dict()
+  else:
+    state_dict = model.state_dict()
+  new_state_dict= {}
+  for k, v in state_dict.items():
+    try:
+      new_state_dict[k] = saved_state_dict[k]
+    except:
+      logger.info("%s is not in the checkpoint" % k)
+      new_state_dict[k] = v
+  if hasattr(model, 'module'):
+    model.module.load_state_dict(new_state_dict)
+  else:
+    model.load_state_dict(new_state_dict)
+  logger.info("Loaded checkpoint '{}' (iteration {})" .format(
+    checkpoint_path, iteration))
+  return model, optimizer, learning_rate, iteration
+
+
+def save_checkpoint(model, optimizer, learning_rate, iteration, checkpoint_path):
+  logger.info("Saving model and optimizer state at iteration {} to {}".format(
+    iteration, checkpoint_path))
+  if hasattr(model, 'module'):
+    state_dict = model.module.state_dict()
+  else:
+    state_dict = model.state_dict()
+  torch.save({'model': state_dict,
+              'iteration': iteration,
+              'optimizer': optimizer.state_dict(),
+              'learning_rate': learning_rate}, checkpoint_path)
+
+
+def summarize(writer, global_step, scalars={}, histograms={}, images={}, audios={}, audio_sampling_rate=22050):
+  for k, v in scalars.items():
+    writer.add_scalar(k, v, global_step)
+  for k, v in histograms.items():
+    writer.add_histogram(k, v, global_step)
+  for k, v in images.items():
+    writer.add_image(k, v, global_step, dataformats='HWC')
+  for k, v in audios.items():
+    writer.add_audio(k, v, global_step, audio_sampling_rate)
+
+
+def latest_checkpoint_path(dir_path, regex="G_*.pth"):
+  f_list = glob.glob(os.path.join(dir_path, regex))
+  f_list.sort(key=lambda f: int("".join(filter(str.isdigit, f))))
+  x = f_list[-1]
+  print(x)
+  return x
+
+
+def plot_spectrogram_to_numpy(spectrogram):
+  global MATPLOTLIB_FLAG
+  if not MATPLOTLIB_FLAG:
+    import matplotlib
+    matplotlib.use("Agg")
+    MATPLOTLIB_FLAG = True
+    mpl_logger = logging.getLogger('matplotlib')
+    mpl_logger.setLevel(logging.WARNING)
+  import matplotlib.pylab as plt
+  import numpy as np
+  
+  fig, ax = plt.subplots(figsize=(10,2))
+  im = ax.imshow(spectrogram, aspect="auto", origin="lower",
+                  interpolation='none')
+  plt.colorbar(im, ax=ax)
+  plt.xlabel("Frames")
+  plt.ylabel("Channels")
+  plt.tight_layout()
+
+  fig.canvas.draw()
+  data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')
+  data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))
+  plt.close()
+  return data
+
+
+def plot_alignment_to_numpy(alignment, info=None):
+  global MATPLOTLIB_FLAG
+  if not MATPLOTLIB_FLAG:
+    import matplotlib
+    matplotlib.use("Agg")
+    MATPLOTLIB_FLAG = True
+    mpl_logger = logging.getLogger('matplotlib')
+    mpl_logger.setLevel(logging.WARNING)
+  import matplotlib.pylab as plt
+  import numpy as np
+
+  fig, ax = plt.subplots(figsize=(6, 4))
+  im = ax.imshow(alignment.transpose(), aspect='auto', origin='lower',
+                  interpolation='none')
+  fig.colorbar(im, ax=ax)
+  xlabel = 'Decoder timestep'
+  if info is not None:
+      xlabel += '\n\n' + info
+  plt.xlabel(xlabel)
+  plt.ylabel('Encoder timestep')
+  plt.tight_layout()
+
+  fig.canvas.draw()
+  data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')
+  data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))
+  plt.close()
+  return data
+
+
+def load_wav_to_torch(full_path):
+  sampling_rate, data = read(full_path)
+  return torch.FloatTensor(data.astype(np.float32)), sampling_rate
+
+
+def load_filepaths_and_text(filename, split="|"):
+  with open(filename, encoding='utf-8') as f:
+    filepaths_and_text = [line.strip().split(split) for line in f]
+  return filepaths_and_text
+
+
+def get_hparams(init=True):
+  parser = argparse.ArgumentParser()
+  parser.add_argument('-c', '--config', type=str, default="./configs/base.json",
+                      help='JSON file for configuration')
+  parser.add_argument('-m', '--model', type=str, required=True,
+                      help='Model name')
+  
+  args = parser.parse_args()
+  model_dir = os.path.join("../drive/MyDrive", args.model)
+
+  if not os.path.exists(model_dir):
+    os.makedirs(model_dir)
+
+  config_path = args.config
+  config_save_path = os.path.join(model_dir, "config.json")
+  if init:
+    with open(config_path, "r") as f:
+      data = f.read()
+    with open(config_save_path, "w") as f:
+      f.write(data)
+  else:
+    with open(config_save_path, "r") as f:
+      data = f.read()
+  config = json.loads(data)
+  
+  hparams = HParams(**config)
+  hparams.model_dir = model_dir
+  return hparams
+
+
+def get_hparams_from_dir(model_dir):
+  config_save_path = os.path.join(model_dir, "config.json")
+  with open(config_save_path, "r") as f:
+    data = f.read()
+  config = json.loads(data)
+
+  hparams =HParams(**config)
+  hparams.model_dir = model_dir
+  return hparams
+
+
+def get_hparams_from_file(config_path):
+  with open(config_path, "r") as f:
+    data = f.read()
+  config = json.loads(data)
+
+  hparams =HParams(**config)
+  return hparams
+
+
+def check_git_hash(model_dir):
+  source_dir = os.path.dirname(os.path.realpath(__file__))
+  if not os.path.exists(os.path.join(source_dir, ".git")):
+    logger.warn("{} is not a git repository, therefore hash value comparison will be ignored.".format(
+      source_dir
+    ))
+    return
+
+  cur_hash = subprocess.getoutput("git rev-parse HEAD")
+
+  path = os.path.join(model_dir, "githash")
+  if os.path.exists(path):
+    saved_hash = open(path).read()
+    if saved_hash != cur_hash:
+      logger.warn("git hash values are different. {}(saved) != {}(current)".format(
+        saved_hash[:8], cur_hash[:8]))
+  else:
+    open(path, "w").write(cur_hash)
+
+
+def get_logger(model_dir, filename="train.log"):
+  global logger
+  logger = logging.getLogger(os.path.basename(model_dir))
+  logger.setLevel(logging.DEBUG)
+  
+  formatter = logging.Formatter("%(asctime)s\t%(name)s\t%(levelname)s\t%(message)s")
+  if not os.path.exists(model_dir):
+    os.makedirs(model_dir)
+  h = logging.FileHandler(os.path.join(model_dir, filename))
+  h.setLevel(logging.DEBUG)
+  h.setFormatter(formatter)
+  logger.addHandler(h)
+  return logger
+
+
+class HParams():
+  def __init__(self, **kwargs):
+    for k, v in kwargs.items():
+      if type(v) == dict:
+        v = HParams(**v)
+      self[k] = v
+    
+  def keys(self):
+    return self.__dict__.keys()
+
+  def items(self):
+    return self.__dict__.items()
+
+  def values(self):
+    return self.__dict__.values()
+
+  def __len__(self):
+    return len(self.__dict__)
+
+  def __getitem__(self, key):
+    return getattr(self, key)
+
+  def __setitem__(self, key, value):
+    return setattr(self, key, value)
+
+  def __contains__(self, key):
+    return key in self.__dict__
+
+  def __repr__(self):
+    return self.__dict__.__repr__()

vits.html

@@ -0,0 +1,12 @@
+<!DOCTYPE NETSCAPE-Bookmark-file-1>
+<!-- This is an automatically generated file.
+It will be read and overwritten.
+DO NOT EDIT! -->
+<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=UTF-8">
+<TITLE>Shortcut</TITLE>
+<DL><p>
+  <DT><H3>vits</H3>
+  <DL><p>
+    <DT><A HREF="https://drive.google.com/open?id=1g8gRN-rfOFkLFXe_ycx4747D4dtnJPTz" ADD_DATE="1660636588">vits</A>
+  </DL><p>
+</DL><p>