Update app.py

app.py

@@ -1,51 +1,46 @@
 import gradio as gr
+import matplotlib.pyplot as plt
+import IPython.display as ipd
+import os
+import json
+import math
 import torch
+from torch import nn
+from torch.nn import functional as F
+from torch.utils.data import DataLoader
 import commons
 import utils
-import os
+from data_utils import TextAudioSpeakerLoader, TextAudioSpeakerCollate
 from models import SynthesizerTrn
 from text.symbols import symbols
 from text import text_to_sequence
+from scipy.io.wavfile import write
 import numpy as np
 
+# 加载情感字典
+emotion_dict = json.load(open("configs/leo.json", "r"))
 
-def get_text(text, hps):
-    text_norm = text_to_sequence(text, hps.data.text_cleaners)
-    if hps.data.add_blank:
-        text_norm = commons.intersperse(text_norm, 0)
-    text_norm = torch.LongTensor(text_norm)
-    return text_norm
-hps = utils.get_hparams_from_file("./configs/vtubers.json")
-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)
+# 加载预训练模型
+hps = utils.get_hparams_from_file("./configs/leo.json")
+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)
 _ = net_g.eval()
+_ = utils.load_checkpoint("logs/leo/G_4000.pth", net_g, None)
 
-_ = utils.load_checkpoint("./nene_final.pth", net_g, None)
-file_names = os.listdir("./wavs")
-
-# 构建新的emotion
-emotion = {}
-for file_name in file_names:
-    emotion_name= os.path.splitext(file_name)[0]
-    emotion[emotion_name] = os.path.join("./wavs", file_name)
-import random
-def tts(txt, emotion):
+# 定义文本转语音函数
+def tts(txt, emotion, roma=False, length_scale=1):
     if roma:
         stn_tst = get_text_byroma(txt, hps)
     else:
         stn_tst = get_text(txt, hps)
-    randsample = None
+    
     with torch.no_grad():
         x_tst = stn_tst.unsqueeze(0)
         x_tst_lengths = torch.LongTensor([stn_tst.size(0)])
         sid = torch.LongTensor([0])
-        if os.path.exists(f"{emotion}.emo.npy"):
-            emo = torch.FloatTensor(np.load(f"{emotion}.emo.npy")).unsqueeze(0)
-        elif emotion == "random_sample":
+        
+        if emotion == "random_sample":
+            # 随机选择一个情感参考音频
+            random_emotion_root = "wavs"
             while True:
                 rand_wav = random.sample(os.listdir(random_emotion_root), 1)[0]
                 if rand_wav.endswith('wav') and os.path.exists(f"{random_emotion_root}/{rand_wav}.emo.npy"):
@@ -53,63 +48,27 @@ def tts(txt, emotion):
             emo = torch.FloatTensor(np.load(f"{random_emotion_root}/{rand_wav}.emo.npy")).unsqueeze(0)
             print(f"{random_emotion_root}/{rand_wav}")
         elif emotion.endswith("wav"):
+            # 从提供的音频中提取情感特征
             import emotion_extract
             emo = torch.FloatTensor(emotion_extract.extract_wav(emotion))
         else:
             print("emotion参数不正确")
+    
+        audio = net_g.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=0.667, noise_scale_w=0.8, length_scale=1.2, emo=emo)[0][0, 0].data.float().numpy()
+    
+    ipd.display(ipd.Audio(audio, rate=hps.data.sampling_rate, normalize=False))
 
-        audio = net_g.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=0.667, noise_scale_w=0.8, length_scale=1, emo=emo)[0][0,0].data.float().numpy()
-    return audio, randsample
-
-
-def tts1(text, emotion):
-    if len(text) > 150:
-        return "Error: Text is too long", None
-    audio, _ = tts(text, emotion)
-    return "Success", (hps.data.sampling_rate, audio)
-
-def tts2(text):
-    if len(text) > 150:
-        return "Error: Text is too long", None
-    audio, randsample = tts(text, "random_sample")
-
-    return str(randsample), (hps.data.sampling_rate, audio)
-
-def tts3(text, emotion):
-    if len(text) > 150:
-        return "Error: Text is too long", None
-    try:
-        audio, _ = tts(text, int(sample))
-        return "Success", (hps.data.sampling_rate, audio)
-    except:
-        return "输入参数不为整数或其他错误", None
-app = gr.Blocks()
-with app:
-    with gr.Tabs():
- #       with gr.TabItem("使用预制情感合成"):
- #           tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
- #           tts_input2 = gr.Dropdown(label="情感", choices=list(emotion_dict.keys()),  value="平静1")
- #           tts_submit = gr.Button("合成音频", variant="primary")
- #           tts_output1 = gr.Textbox(label="Message")
- #           tts_output2 = gr.Audio(label="Output")
- #           tts_submit.click(tts1, [tts_input1, tts_input2], [tts_output1, tts_output2])
- #       with gr.TabItem("随机抽取训练集样本作为情感参数"):
- #           tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
- #           tts_submit = gr.Button("合成音频", variant="primary")
- #           tts_output1 = gr.Textbox(label="随机样本id（可用于第三个tab中合成）")
- #           tts_output2 = gr.Audio(label="Output")
- #           tts_submit.click(tts2, [tts_input1], [tts_output1, tts_output2])
+# 创建GUI界面
+def run_tts(text, emotion, roma=False):
+    tts(text, emotion, roma)
 
-        with gr.TabItem("使用指定情感样本作为情感参数"):
+inputs = [
+    gr.inputs.Textbox(label="请输入文本"),
+    gr.inputs.Textbox(label="请输入参考音频路径或选择'random_sample'随机选择"),
+    gr.inputs.Checkbox(label="是否使用音素合成")
+]
 
-            tts_input1 = gr.TextArea(label="日语文本", value="こんにちは。私わあやちねねです。")
-            tts_input2 = gr.TextArea(label="情感样本名字", value="")
-            tts_submit = gr.Button("合成音频", variant="primary")
-            tts_output1 = gr.Textbox(label="Message")
-            tts_output2 = gr.Audio(label="Output")
-            tts_submit.click(tts3, [tts_input1, tts_input2], [tts_output1, tts_output2])
+outputs = gr.outputs.Audio(label="合成音频")
 
-        with gr.TabItem("使用参考音频作为情感参数"):
-            tts_input1 = gr.TextArea(label="text", value="暂未实现")
+interface = gr.Interface(fn=run_tts, inputs=inputs, outputs=outputs, title="中文文本转
 
-    app.launch()