app.py

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 gradio as gr


pth_path = "G_240000.pth"
hps = utils.get_hparams_from_file("./configs/hoshimi_base.json")
# device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
device = torch.device("cpu")

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 load_model(pth_path):
    net_g = SynthesizerTrn(
        len(symbols),
        hps.data.filter_length // 2 + 1,
        hps.train.segment_size // hps.data.hop_length,
        **hps.model).to(device)
    _ = net_g.eval()

    _ = utils.load_checkpoint(pth_path, net_g, None)
    return net_g


def list_model():
    global pth_path
    res = []
    dir = os.getcwd()
    for f in os.listdir(dir):
        if (f.startswith("D_")):
            continue
        if (f.endswith(".pth")):
            res.append(f)
            if len(f) >= len(pth_path):
                pth_path = f
    return res


def infer(text):
    stn_tst = get_text(text, hps)
    with torch.no_grad():
        x_tst = stn_tst.unsqueeze(0).to(device)
        x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).to(device)
        audio = net_g.infer(x_tst, x_tst_lengths, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][0,0].data.float().numpy()
    return (hps.data.sampling_rate, audio)


models = list_model()
net_g = load_model(pth_path)

def change_model(model):
    global pth_path
    global net_g_ms
    pth_path = model
    net_g_ms = load_model(pth_path)
    return "载入模型："+pth_path


app = gr.Blocks()
with app:
    with open("header.html", "r") as f:
        gr.HTML(f.read())
    with gr.Tabs():
        with gr.TabItem("Basic"):
            choice_model = gr.Dropdown(
                choices=models, label="模型", value=pth_path)
            tts_input1 = gr.TextArea(
                label="请输入文本（目前只支持汉字和单个英文字母，也可以使用逗号、句号、感叹号、空格等常用符号来改变语调和停顿）",
                value="这里是爱喝奶茶，穿得也像奶茶魅力点是普通话二乙的星弥吼西咪，晚上齁。")
            tts_submit = gr.Button("用文本合成", variant="primary")
            tts_output = gr.Audio(label="Output")
            tts_model = gr.Markdown("")
            tts_submit.click(infer, [tts_input1], [tts_output])
            choice_model.change(change_model, inputs=[
                                choice_model], outputs=[tts_model])
    app.launch()