app.py

import time
import matplotlib.pyplot as plt
import IPython.display as ipd
import re
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 gradio as gr
import commons
import utils
from data_utils import TextAudioLoader, TextAudioCollate, TextAudioSpeakerLoader, TextAudioSpeakerCollate
from models import SynthesizerTrn
from text.symbols import symbols
from text import text_to_sequence
import unicodedata
from scipy.io.wavfile import write
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


def get_label(text, label):
    if f'[{label}]' in text:
        return True, text.replace(f'[{label}]', '')
    else:
        return False, text





def selection(speaker):
    if speaker == "高咲侑":
        spk = 0
        return spk

    elif speaker == "歩夢":
        spk = 1
        return spk

    elif speaker == "かすみ":
        spk = 2
        return spk

    elif speaker == "しずく":
        spk = 3
        return spk

    elif speaker == "果林":
        spk = 4
        return spk
    
    elif speaker == "愛":
        spk = 5
        return spk

    elif speaker == "彼方":
        spk = 6
        return spk

    elif speaker == "せつ菜":
        spk = 7
        return spk
    elif speaker == "エマ":
        spk = 8
        return spk
    elif speaker == "璃奈":
        spk = 9
        return spk
    elif speaker == "栞子":
        spk = 10
        return spk
    elif speaker == "ランジュ":
        spk = 11
        return spk
    elif speaker == "ミア":
        spk = 12
        return spk
    elif speaker == "派蒙":
        spk = 16
        return spk
    
def sle(language,tts_input0):
    if language == "中文":
        tts_input1 = "[ZH]" + tts_input0.replace('\n','。').replace(' ',',') + "[ZH]"
        return tts_input1
    if language == "英文":
        tts_input1 = "[EN]" + tts_input0.replace('\n','.').replace(' ',',') + "[EN]"
        return tts_input1
    elif language == "日文":
        tts_input1 = "[JA]" + tts_input0.replace('\n','。').replace(' ',',') + "[JA]"
        return tts_input1
def infer(language,text,speaker_id, n_scale= 0.667,n_scale_w = 0.8, l_scale = 1 ):
    speaker_id = int(selection(speaker_id))
    answer = sle(language,text)
    stn_tst = get_text(answer, hps_ms)
    with torch.no_grad():
        x_tst = stn_tst.unsqueeze(0).to(dev)
        x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).to(dev)
        sid = torch.LongTensor([speaker_id]).to(dev)
        t1 = time.time()
        audio = net_g_ms.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=n_scale, noise_scale_w=n_scale_w, length_scale=l_scale)[0][0,0].data.cpu().float().numpy()
        t2 = time.time()
        spending_time = "推理时间："+str(t2-t1)+"s"
        image = '1.png'
        print(spending_time)
    return (hps_ms.data.sampling_rate, audio),image
lan = ["中文","日文","英文"]
idols = ["高咲侑","歩夢","かすみ","しずく","果林","愛","彼方","せつ菜","璃奈","栞子","エマ","ランジュ","ミア","派蒙"]
dev = torch.device("cpu")
hps_ms = utils.get_hparams_from_file("config.json")
net_g_ms = SynthesizerTrn(
    len(symbols),
    hps_ms.data.filter_length // 2 + 1,
    hps_ms.train.segment_size // hps_ms.data.hop_length,
    n_speakers=hps_ms.data.n_speakers,
    **hps_ms.model).to(dev)
_ = net_g_ms.eval()

_ = utils.load_checkpoint("G_874000.pth", net_g_ms, None)

app = gr.Blocks()

with app:
    with gr.Tabs():
        with gr.TabItem("Basic"):
            tts_input1 = gr.TextArea(label="VITS模型，绝赞训练中", value="一次審査、二次審査、それぞれの欄に記入をお願いします。")
            language = gr.Dropdown(label="选择语言",choices=lan, value="日文", interactive=True)
            para_input1 = gr.Slider(minimum= 0.01,maximum=1.0,label="更改噪声比例", value=0.667)
            para_input2 = gr.Slider(minimum= 0.01,maximum=1.0,label="更改噪声偏差", value=0.8)
            para_input3 = gr.Slider(minimum= 0.1,maximum=10,label="更改时间比例", value=1)
            tts_submit = gr.Button("Generate", variant="primary")
            speaker1 = gr.Dropdown(label="选择说话人",choices=idols, value="かすみ", interactive=True)
            tts_output2 = gr.Audio(label="Output")
            tts_output3 = gr.Image(label = "Model")
            tts_submit.click(infer, [language,tts_input1,speaker1,para_input1,para_input2,para_input3], [tts_output2,tts_output3])
    #app.launch(share=True)
    app.launch()