TTSs/genshin_api_tts.py

import io
import json
import os

import gradio as gr
import requests
from pydub import AudioSegment

from TTSs.base_tts import Base_TTS


class genshin_api_TTS(Base_TTS):
    def __init__(self):
        self.genshin_api = "https://tirs.ai-lab.top"
        self.speakers_genshin = self.get_spk()

    def get_name(self):
        return '原神语音合成-api'

    def _get_submit_button(self):
        """
        要求返回1个参数：btn
        btn: gr.Button对象，是提交按钮

        :return: btn
        """
        btn = gr.Button(value="原神提交", variant="primary", interactive=True, visible=False)
        return btn

    def _get_config_page(self):
        with gr.Group(visible=False) as config_genshin_api:
            gr.Markdown('原作者：https://modelscope.cn/studios/erythrocyte/Bert-VITS2_Genshin_TTS/summary')
            with gr.Row():
                token = gr.Textbox(label="AccessToken", placeholder="请输入AccessToken",
                                   type="password",
                                   interactive=True, value=os.environ.get('GENSHIN_API_KEY', ''))
                appid = gr.Textbox(label="AppID", placeholder="请输入AppID", interactive=True,
                                   value='ig8t76x6036h3hpw')
            with gr.Row():
                gr.Markdown(
                    "AppID不要改，AccessToken最好换自己的。")

            with gr.Row():
                speaker = gr.Dropdown(choices=self.speakers_genshin,
                                      value=self.speakers_genshin[0], label="角色",
                                      interactive=True)
                search = gr.Textbox(label="搜索角色", lines=1, interactive=True)
                btn2 = gr.Button(value="搜索")
            with gr.Column():
                with gr.Row():
                    sdp_ratio = gr.Slider(minimum=0, maximum=1, value=0.2, step=0.1, label="SDP/DP 混合比",
                                          interactive=True)
                    noise_scale = gr.Slider(minimum=0.1, maximum=2, value=0.6, step=0.1, label="感情",
                                            interactive=True)
                with gr.Row():
                    noise_scale_w = gr.Slider(minimum=0.1, maximum=2, value=0.8, step=0.1, label="音素长度",
                                              interactive=True)
                    length_scale = gr.Slider(minimum=-99, maximum=99, value=0, step=0.1, label="语速(%)",
                                             interactive=True)

            btn2.click(self.search_speaker, inputs=[search], outputs=[speaker])

        inputs = [
            token, appid, speaker, sdp_ratio, noise_scale, noise_scale_w,
            length_scale
        ]

        return config_genshin_api, inputs

    def _generate(self, text, token, appid, speaker, sdp, noise, noisew, length):
        speed = (100 - length) / 100
        headers = {'Content-Type': 'application/json'}
        infer_info = {'lang': 'zh', 'appid': appid, 'token': token, 'speaker': speaker, 'text': text,
                      'sdp_ratio': sdp, 'noise': noise, 'noisew': noisew, 'length': speed}
        resp = requests.post(url=f"{self.genshin_api}/api/ex/vits", headers=headers,
                             data=json.dumps(infer_info))
        data = json.loads(resp.text)

        wmv_file = requests.get(data["audio"]).content

        original_audio = AudioSegment.from_wav(io.BytesIO(wmv_file))
        return original_audio

    def get_spk(self):
        resp = requests.get(url=f"{self.genshin_api}/spklist/spks.json")
        data = json.loads(resp.text)
        return data

    def search_speaker(self, search_value):
        for s in self.speakers_genshin:
            if search_value == s:
                return s
        for s in self.speakers_genshin:
            if search_value in s:
                return s