app.py

import logging
import re

import gradio as gr
import numpy
import torch

import utils
from infer import infer, get_net_g

logging.getLogger("numba").setLevel(logging.WARNING)
logging.getLogger("markdown_it").setLevel(logging.WARNING)
logging.getLogger("urllib3").setLevel(logging.WARNING)
logging.getLogger("matplotlib").setLevel(logging.WARNING)

logging.basicConfig(level=logging.INFO, format="| %(name)s | %(levelname)s | %(message)s")

logger = logging.getLogger(__name__)

net_g = None
hps = None

device = "cuda" if torch.cuda.is_available() else "cpu"
model_path = "models/G_1000.pth"
sampling_rate = 22050


def split_sentence(sentence: str):
    if len(sentence) == 0:
        return []

    result = []

    is_english = [i.isascii() for i in sentence]
    is_chinese = [not re.match(r"[a-zA-Z]", i) for i in sentence]

    assert len(is_english) == len(is_chinese) == len(sentence), "bad length"
    assert is_english[0] or is_chinese[0], "bad first char: " + sentence[0]

    current_language = ''
    current_chain = []
    for idx in range(len(sentence)):
        if not is_english[idx]:
            current_language = 'ZH'
            current_chain = is_chinese
            break
        if not is_chinese[idx]:
            current_language = 'EN'
            current_chain = is_english
            break
        pass

    step = 0
    while step < len(sentence):
        try:
            next_step = current_chain.index(False, step)
        except ValueError:
            next_step = len(sentence)
        result.append((sentence[step:next_step], current_language))
        step = next_step
        current_language = 'ZH' if current_language == 'EN' else 'EN'
        current_chain = is_chinese if current_language == 'ZH' else is_english
        pass

    return result


def tts_fn(
        text: str,
        speaker,
        sdp_ratio,
        noise_scale,
        noise_scale_w,
        length_scale,
        language,
):
    language = 'ZH' if language == '普通话' else 'SH'
    sentences = split_sentence(text)

    silence = numpy.zeros(sampling_rate // 2, dtype=numpy.int16)
    audio_data = numpy.array([], dtype=numpy.float32)
    for (sentence, sentence_language) in sentences:
        sub_audio_data = infer(
            sentence,
            sdp_ratio,
            noise_scale,
            noise_scale_w,
            length_scale,
            sid=speaker,
            language=language if sentence_language == "ZH" else sentence_language,
            hps=hps,
            net_g=net_g,
            device=device)
        audio_data = numpy.concatenate((audio_data, sub_audio_data, silence))

    audio_data = audio_data / numpy.abs(audio_data).max()
    audio_data = audio_data * 32767
    audio_data = audio_data.astype(numpy.int16)

    return "Success", (sampling_rate, audio_data)


def main():
    logging.basicConfig(level=logging.DEBUG)

    global hps
    hps = utils.get_hparams_from_file("configs/config.json")

    global net_g
    net_g = get_net_g(model_path=model_path, device=device, hps=hps)

    speaker_ids = hps.data.spk2id
    speakers = list(speaker_ids.keys())
    languages = ["上海话", "普通话"]
    with gr.Blocks() as app:
        with gr.Row():
            with gr.Column():
                gr.Markdown('# Bert-VITS2-Shanghainese')
        with gr.Row():
            with gr.Column():
                text = gr.TextArea(
                    label="输入文本内容",
                    value="\n".join([
                        "站一个制高点看上海，",
                        "Looking at Shanghai from a commanding height,",
                        "上海的弄堂是壮观的景象。",
                        "The alleys in Shanghai are a great sight.",
                        "它是这城市背景一样的东西。",
                        "It is something with the same background as this city."
                    ]),
                )
                sdp_ratio = gr.Slider(minimum=0, maximum=1, value=0.2, step=0.1, label="SDP/DP混合比")
                noise_scale = gr.Slider(minimum=0.1, maximum=2, value=0.6, step=0.1, label="感情")
                noise_scale_w = gr.Slider(minimum=0.1, maximum=2, value=0.8, step=0.1, label="音素长度")
                length_scale = gr.Slider(minimum=0.1, maximum=2, value=1.0, step=0.1, label="语速")
            with gr.Column():
                with gr.Row():
                    with gr.Column():
                        speaker = gr.Dropdown(choices=speakers, value=speakers[0], label="选择说话人")
                    with gr.Column():
                        language = gr.Dropdown(choices=languages, value=languages[0], label="选择语言")
                submit_btn = gr.Button("生成音频", variant="primary")
                text_output = gr.Textbox(label="状态")
                audio_output = gr.Audio(label="音频")
        with gr.Row():
            with gr.Column():
                gr.Markdown('### Thanks <a href="https://github.com/fishaudio/Bert-VITS2" target="_blank">github@Bert-VITS2</a>')
        submit_btn.click(
            tts_fn,
            inputs=[
                text,
                speaker,
                sdp_ratio,
                noise_scale,
                noise_scale_w,
                length_scale,
                language,
            ],
            outputs=[text_output, audio_output],
        )

    app.launch(share=False, server_name="0.0.0.0", server_port=7860)


if __name__ == "__main__":
    main()