app.py

import gradio as gr
# import matplotlib.pyplot as plt
import logging
# logger = logging.getLogger(__name__)
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
from data_utils import TextAudioLoader, TextAudioCollate, TextAudioSpeakerLoader, TextAudioSpeakerCollate
from models import SynthesizerTrn
from text.symbols import symbols
from text import text_to_sequence
import time


def get_text(text, hps):
    # text_norm = requests.post("http://121.5.171.42:39001/texttosequence?text="+text).json()["text_norm"]
    text_norm = text_to_sequence(text, hps.data.text_cleaners)
    # print(hps.data.text_cleaners)
    # print(text_norm)
    if hps.data.add_blank:
        text_norm = commons.intersperse(text_norm, 0)
    text_norm = torch.LongTensor(text_norm)
    return text_norm

def load_model(config_path, pth_path):
    global dev, hps, net_g
    dev = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
    hps = utils.get_hparams_from_file(config_path)

    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).to(dev)
    _ = net_g.eval()
    
    _ = utils.load_checkpoint(pth_path, net_g)

    print(f"{pth_path}加载成功!")
    
def infer(c_name, text):
    c_id = character_dict[c_name]
    stn_tst = get_text(text, hps)
    with torch.no_grad():
        x_tst = stn_tst.to(dev).unsqueeze(0)
        x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).to(dev)
        sid = torch.LongTensor([c_id]).to(dev)
        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()

        return (hps.data.sampling_rate, audio)

pth_path = "model/G_215000.pth"
config_path = "configs/config.json"
character_dict = {
    "夜刀神十香": 1,
    "鸢一折纸": 2,
    "时崎狂三": 3,
    "冰芽川四糸乃": 4,
    "五河琴里": 5,
    "八舞夕弦": 6,
    "八舞耶俱矢": 7,
    "诱宵美九": 8,
    "园神凛祢": 9,
    "园神凛绪": 11,
    "或守鞠亚": 12,
    "或守鞠奈": 13,
    "崇宫真那": 14,
}


load_model(config_path, pth_path)

app = gr.Blocks()
with app:
    with gr.Tabs():
        with gr.Row():
            text = gr.TextArea(
                label="请输入文本(仅支持日语)", value="こんにちは,世界!")
        with gr.Row():
            radio = gr.Radio(list(character_dict.keys()), 
                label="请选择角色")
        with gr.Row():
            tts_submit = gr.Button("合成", variant="primary")
        with gr.Row():
            tts_output = gr.Audio(label="Output")
        # model_submit.click(load_model, [config_path, pth_path])
        tts_submit.click(infer, [radio, text], [tts_output])
        radio.change(infer, [radio, text], [tts_output])

    gr.HTML("""
    <div style="text-align:center">
                    <h4 class="h-sign" style="font-size: 12px;">
                        这是一个使用<a href="https://github.com/thesupersonic16/DALTools" target="_blank">thesupersonic16/DALTools</a>提供的解包音频作为数据集，
                        使用<a href="https://github.com/jaywalnut310/vits" target="_blank">VITS</a>技术训练的语音合成demo(215k step)。
                    </h4>
                </div>
            </div>
<div style="text-align:center"> 
    仅供学习交流，不可用于商业或非法用途
    <br/>
    使用本项目模型直接或间接生成的音频，必须声明由AI技术或VITS技术合成
</div>
    """)
    app.launch()