VirtualWife增加火山引擎语音复刻

2025-09-16

折腾

暂无评论

116 次阅读

VirtualWife是一个开源的虚拟数字人项目。其中数字人生成的部分是基于ChatVRM的虚拟主播项目，语音是由多个TTS驱动，文字生成可以使用LLM。

yakami129/VirtualWife: VirtualWife是一个虚拟数字人项目，支持B站直播，支持openai、ollama

pixiv/ChatVRM

增加domain-chatbot/apps/speech/tts/bytedance_tts.py文件：

import json
import os
import base64
import requests
from ..utils.uuid_generator import generate

url = "https://openspeech.bytedance.com/api/v1/tts"
access_token = "Nxxx-WQfixxxxx-xxx"
headers = {
    "Authorization": f"Bearer;{access_token}"
}

bytedance_tts_voices = [{
    "id": "S_xxxxx",
    "name": "xxx"
}]


class BytedanceTTSAPI:
    def request(self, params: dict[str, str]) -> str:
        # 合成语音
        body = json.dumps(params, ensure_ascii=False).encode('utf-8')
        response = requests.post(url, headers=headers, data=body, verify=False)

        # 初始化文件夹
        file_name = generate() + ".wav"
        file_path = os.getcwd() + "/tmp/" + file_name
        dirPath = os.path.dirname(file_path)
        if not os.path.exists(dirPath):
            os.makedirs(dirPath)
        if not os.path.exists(file_path):
            # 用open创建文件 兼容mac
            open(file_path, 'a').close()

        # 写入语音文件
        with open(file_path, 'wb') as file:
            # file.write(response.content)
            data = json.loads(response.text)["data"]
            file.write(base64.b64decode(data))
        return file_name


class Bytedance:
    client: BytedanceTTSAPI

    def __init__(self):
        self.client = BytedanceTTSAPI()

    def synthesis(self, text: str, speaker: str) -> str:
        params = {
            # "data": [text, speaker, sdp_ratio, noise, noisew, 1, "ZH", False, 1, 0.2, None, "Happy", "", 0.7],
            # "event_data": None,
            # "fn_index": 0,
            # "session_hash": str(uuid.uuid4())
            "app": {
                "appid": "xxxxx",
                "token": "NxHV-xxx-xxx",
                "cluster": "volcano_icl"
            },
            "user": {
                "uid": "uid123"
            },
            "audio": {
                "voice_type": "S_xxxxxx",
                "encoding": "wav",
                "speed_ratio": 1
            },
            "request": {
                "reqid": "uuid",
                "text": text,
                "operation": "query"
            }
        }
        return self.client.request(params=params)

    def get_voices(self) -> list:
        return bytedance_tts_voices


if __name__ == '__main__':
    client = Bytedance()
    client.synthesis(text="你好", speaker="流萤_ZH")

本质上就是HTTP请求生成语音即可。

domain-chatbot/apps/speech/tts/tts_driver.py中增加：

class BytedanceTTS(BaseTTS):
    '''ByteDance 语音合成类'''
    client: Bytedance

    def __init__(self):
        self.client = Bytedance()

    def synthesis(self, text: str, voice_id: str, **kwargs) -> str:
        return self.client.synthesis(text=text, speaker=voice_id)

    def get_voices(self) -> list[dict[str, str]]:
        return self.client.get_voices()