├── src ├── api │ ├── requirements.txt │ ├── .env.example │ └── app │ │ ├── utils.py │ │ ├── tts_handler.py │ │ └── server.py ├── setup.sh ├── openai-edge-tts.sh ├── startup.sh └── sshx.sh ├── 免费部署微软EdgeTTS,适配openai语音api接口，一键部署教程-封面.jpg ├── Dockerfile └── README.md /src/api/requirements.txt: -------------------------------------------------------------------------------- 1 | flask 2 | gevent 3 | python-dotenv 4 | edge-tts 5 | art 6 | -------------------------------------------------------------------------------- /免费部署微软EdgeTTS,适配openai语音api接口，一键部署教程-封面.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/aigem/edgeTTS-openai-api/HEAD/免费部署微软EdgeTTS,适配openai语音api接口，一键部署教程-封面.jpg -------------------------------------------------------------------------------- /src/setup.sh: -------------------------------------------------------------------------------- 1 | #!/bin/sh 2 | 3 | # 在这里添加任何必要的设置步骤 4 | echo "正在进行设置..." 5 | 6 | # 安装 pnpm pm2 7 | npm install -g pnpm pm2 8 | 9 | echo "设置完成" 10 | -------------------------------------------------------------------------------- /src/api/.env.example: -------------------------------------------------------------------------------- 1 | API_KEY=your_api_key_here 2 | PORT=7860 3 | 4 | DEFAULT_VOICE=en-US-AndrewNeural 5 | DEFAULT_RESPONSE_FORMAT=mp3 6 | DEFAULT_SPEED=1.0 7 | 8 | DEFAULT_LANGUAGE=zh-CN 9 | 10 | REQUIRE_API_KEY=True -------------------------------------------------------------------------------- /src/openai-edge-tts.sh: -------------------------------------------------------------------------------- 1 | #!/bin/sh 2 | set -e 3 | 4 | echo "开始安装 openai-edge-tts..." 5 | 6 | # 检查 openai-edge-tts 文件夹是否存在 7 | if [ ! -d "$HOMEDIR/edgeTTS-openai-api/src/api" ]; then 8 | echo "openai-edge-tts 文件夹不存在" 9 | exit 1 10 | fi 11 | 12 | # 进入 openai-edge-tts 文件夹并安装依赖 13 | cd $HOMEDIR/edgeTTS-openai-api/src/api 14 | pip install -r requirements.txt 15 | 16 | # .env 文件 17 | cp .env.example .env 18 | 19 | echo "openai-edge-tts 安装完成" 20 | -------------------------------------------------------------------------------- /src/startup.sh: -------------------------------------------------------------------------------- 1 | #!/bin/sh 2 | 3 | cat $HOMEDIR/edgeTTS-openai-api/src/api/.env 4 | 5 | # 检查 OPENAI_EDGE_TTS_INSTALL是否为true 6 | if [ "$OPENAI_EDGE_TTS_INSTALL" = true ]; then 7 | cd ${HOMEDIR}/edgeTTS-openai-api/src/api 8 | # 先尝试删除已存在的实例 9 | pm2 delete openai-edge-tts 2>/dev/null || true 10 | # 启动新的实例 11 | pm2 start python --name openai-edge-tts -- ${HOMEDIR}/edgeTTS-openai-api/src/api/app/server.py 12 | pm2 save 13 | fi 14 | 15 | cd ${HOMEDIR} 16 | 17 | # 只输出日志 18 | pm2 logs --lines 50 19 | -------------------------------------------------------------------------------- /src/sshx.sh: -------------------------------------------------------------------------------- 1 | #!/bin/sh 2 | set -e 3 | 4 | echo "安装 sshx..." 5 | 6 | # 设置下载URL和目标目录 7 | DOWNLOAD_URL="https://sshx.s3.amazonaws.com/sshx-x86_64-unknown-linux-musl.tar.gz" 8 | TARGET_DIR="$HOMEDIR/sshx" 9 | 10 | # 创建目标目录 11 | mkdir -p "$TARGET_DIR" 12 | chown $USER:$USER "$TARGET_DIR" 13 | 14 | # 下载tar.gz文件 15 | echo "Downloading sshx..." 16 | curl -L "$DOWNLOAD_URL" -o /tmp/sshx.tar.gz 17 | 18 | # 解压文件到目标目录 19 | echo "Extracting sshx..." 20 | tar -xzf /tmp/sshx.tar.gz -C "$TARGET_DIR" 21 | 22 | # 清理临时文件 23 | rm /tmp/sshx.tar.gz 24 | 25 | # 检查是否成功解压 26 | if [ -f "$TARGET_DIR/sshx" ]; then 27 | echo "sshx has been successfully downloaded and extracted to $TARGET_DIR" 28 | chmod +x "$TARGET_DIR/sshx" 29 | echo "sshx 版本: $($TARGET_DIR/sshx -V)" 30 | else 31 | echo "Failed to extract sshx. Please check the download URL and try again." 32 | exit 1 33 | fi 34 | 35 | # 将sshx添加到PATH 36 | export PATH="$PATH:$TARGET_DIR" 37 | 38 | echo "sshx 安装成功" 39 | 40 | -------------------------------------------------------------------------------- /Dockerfile: -------------------------------------------------------------------------------- 1 | FROM nikolaik/python-nodejs:python3.10-nodejs20 2 | 3 | # 添加 CACHEBUST 参数 4 | ARG CACHEBUST=1 5 | 6 | ENV USER=pn \ 7 | HOMEDIR=/home/pn \ 8 | PORT=7860 \ 9 | SSHX_INSTALL=true \ 10 | OPENAI_EDGE_TTS_INSTALL=true 11 | 12 | RUN apt-get update && apt-get install -y --no-install-recommends \ 13 | apt-utils \ 14 | build-essential \ 15 | libpq-dev \ 16 | neofetch \ 17 | git \ 18 | curl \ 19 | vim \ 20 | && rm -rf /var/lib/apt/lists/* 21 | 22 | WORKDIR ${HOMEDIR} 23 | 24 | # 使用 CACHEBUST 参数来强制更新 25 | RUN git clone https://github.com/aigem/edgeTTS-openai-api.git \ 26 | && chmod +x ${HOMEDIR}/edgeTTS-openai-api/src/*.sh \ 27 | && ls -la ${HOMEDIR}/edgeTTS-openai-api/src \ 28 | && ls -la ${HOMEDIR}/edgeTTS-openai-api \ 29 | && ${HOMEDIR}/edgeTTS-openai-api/src/setup.sh \ 30 | && if [ "$SSHX_INSTALL" = true ]; then ${HOMEDIR}/edgeTTS-openai-api/src/sshx.sh; fi \ 31 | && if [ "$OPENAI_EDGE_TTS_INSTALL" = true ]; then ${HOMEDIR}/edgeTTS-openai-api/src/openai-edge-tts.sh; fi 32 | 33 | # 暴露 Remix 端口 34 | EXPOSE ${PORT} 35 | 36 | # 在启动时设置 API_KEY 环境变量 37 | ENTRYPOINT /home/pn/edgeTTS-openai-api/src/startup.sh 38 | -------------------------------------------------------------------------------- /src/api/app/utils.py: -------------------------------------------------------------------------------- 1 | # utils.py 2 | 3 | from flask import request, jsonify 4 | from functools import wraps 5 | import os 6 | from dotenv import load_dotenv 7 | 8 | load_dotenv() 9 | 10 | def getenv_bool(name: str, default: bool = False) -> bool: 11 | return os.getenv(name, str(default)).lower() in ("yes", "y", "true", "1", "t") 12 | 13 | API_KEY = os.getenv('API_KEY', 'your_api_key_here') 14 | REQUIRE_API_KEY = getenv_bool('REQUIRE_API_KEY', True) 15 | 16 | def require_api_key(f): 17 | @wraps(f) 18 | def decorated_function(*args, **kwargs): 19 | if not REQUIRE_API_KEY: 20 | return f(*args, **kwargs) 21 | auth_header = request.headers.get('Authorization') 22 | if not auth_header or not auth_header.startswith('Bearer '): 23 | return jsonify({"error": "Missing or invalid API key"}), 401 24 | token = auth_header.split('Bearer ')[1] 25 | if token != API_KEY: 26 | return jsonify({"error": "Invalid API key"}), 401 27 | return f(*args, **kwargs) 28 | return decorated_function 29 | 30 | # Mapping of audio format to MIME type 31 | AUDIO_FORMAT_MIME_TYPES = { 32 | "mp3": "audio/mpeg", 33 | "opus": "audio/ogg", 34 | "aac": "audio/aac", 35 | "flac": "audio/flac", 36 | "wav": "audio/wav", 37 | "pcm": "audio/L16" 38 | } 39 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # OpenAI 通用的微软 Edge TTS API 2 | 3 | 这个项目提供了一个基于微软 Edge TTS 的 OpenAI TTS API 替代品。它允许用户通过 API 调用将文本转换为语音,支持多种语音和音频格式。 4 | 5 | ## 功能特点 6 | 7 | - 兼容 OpenAI TTS API 的接口 8 | - 支持多种语音和语言 9 | - 可调节语音速度 10 | - 支持多种音频输出格式(mp3, opus, aac, flac, wav, pcm) 11 | - 提供模型和可用语音列表的 API 端点 12 | - 可选的 API 密钥认证 13 | - Huggingface 支持,便于部署 14 | 15 | ## 安装 16 | 17 | ### 方法 1: 18 | 打开 Huggingface 页面，点击 `复制空间` 按钮，直接部署。 19 | 20 | [Huggingface 部署链接](https://huggingface.co/spaces/fuliai/edgetts-api-openai) 21 | 22 | 最简单的方法。 23 | 24 | 视频教程 [B站](https://www.bilibili.com/video/BV1M8yiYNEkj/) 25 | ![img](https://raw.githubusercontent.com/aigem/edgeTTS-openai-api/refs/heads/main/%E5%85%8D%E8%B4%B9%E9%83%A8%E7%BD%B2%E5%BE%AE%E8%BD%AFEdgeTTS%2C%E9%80%82%E9%85%8Dopenai%E8%AF%AD%E9%9F%B3api%E6%8E%A5%E5%8F%A3%EF%BC%8C%E4%B8%80%E9%94%AE%E9%83%A8%E7%BD%B2%E6%95%99%E7%A8%8B-%E5%B0%81%E9%9D%A2.jpg) 26 | 27 | ### 方法 2: 复制Dockerfile文件内容，在Huggingface上新建空间后，新建Dockerfile文件，粘贴内容,保存就`部署`了。 28 | 29 | ### 方法 3: 本地安装 30 | 31 | 1. 克隆仓库: 32 | bash 33 | git clone https://github.com/aigem/edgeTTS-openai-api.git 34 | cd edgeTTS-openai-api/src/api 35 | 36 | 37 | 2. 安装依赖: 38 | 39 | bash 40 | pip install -r requirements.txt 41 | 42 | 3. 配置环境变量: 43 | 44 | 复制 `.env.example` 文件为 `.env`,并根据需要修改其中的设置: 45 | bash 46 | cp .env.example .env 47 | 48 | 2. 服务器将在 http://localhost:PORT 上运行(具体查看 .env 文件) 49 | 50 | ## 使用 API: 51 | 52 | - 文本转语音: POST `/v1/audio/speech` 53 | - 获取可用模型列表: GET/POST `/v1/models` 54 | - 获取可用语音列表: GET/POST `/v1/voices` 55 | - 获取所有可用语音: GET/POST `/v1/voices/all` 56 | 57 | ### API 使用示例 58 | 59 | #### 文本转语音 60 | 61 | bash:README.md 62 | curl http://localhost:5050/v1/audio/speech \ 63 | -H "Authorization: Bearer your_api_key_here" \ 64 | -H "Content-Type: application/json" \ 65 | -d '{ 66 | "model": "tts-1", 67 | "input": "Hello, world!", 68 | "voice": "alloy", 69 | "response_format": "mp3", 70 | "speed": 1.0 71 | }' \ 72 | --output speech.mp3 73 | 74 | 75 | #### 获取可用模型列表 76 | bash 77 | curl http://localhost:5050/v1/models \ 78 | -H "Authorization: Bearer your_api_key_here" 79 | 80 | ## 环境变量 81 | 82 | - `API_KEY`: API 密钥(默认: 'your_api_key_here') 83 | - `PORT`: 服务器端口(默认: 5050) 84 | - `DEFAULT_VOICE`: 默认语音(默认: 'en-US-AndrewNeural') 85 | - `DEFAULT_RESPONSE_FORMAT`: 默认音频格式(默认: 'mp3') 86 | - `DEFAULT_SPEED`: 默认语音速度(默认: 1.0) 87 | - `DEFAULT_LANGUAGE`: 默认语言(默认: 'en-US') 88 | - `REQUIRE_API_KEY`: 是否要求 API 密钥认证(默认: true) 89 | 90 | ## 待办事项 91 | 92 | - 增加 Hugging Face 空间环境变量设置 API_KEY 93 | - 开发用户友好的 Web 界面 94 | 95 | ## 贡献 96 | 97 | 欢迎提交问题报告和拉取请求。对于重大更改,请先开issue讨论您想要改变的内容。 98 | 99 | ## 许可证 100 | 101 | [MIT](https://choosealicense.com/licenses/mit/) 102 | 103 | ## 相关链接 104 | 105 | - [视频教程](https://www.youtube.com/@all.ai.) 106 | - [项目 GitHub 仓库](https://github.com/aigem/edgeTTS-openai-api) 107 | -------------------------------------------------------------------------------- /src/api/app/tts_handler.py: -------------------------------------------------------------------------------- 1 | # tts_handler.py 2 | 3 | import edge_tts 4 | import asyncio 5 | import tempfile 6 | import subprocess 7 | import os 8 | 9 | # Language default (environment variable) 10 | DEFAULT_LANGUAGE = os.getenv('DEFAULT_LANGUAGE', 'en-US') 11 | 12 | # OpenAI voice names mapped to edge-tts equivalents 13 | voice_mapping = { 14 | 'alloy': 'en-US-AvaNeural', 15 | 'echo': 'en-US-AndrewNeural', 16 | 'fable': 'en-GB-SoniaNeural', 17 | 'onyx': 'en-US-EricNeural', 18 | 'nova': 'en-US-SteffanNeural', 19 | 'shimmer': 'en-US-EmmaNeural' 20 | } 21 | 22 | async def _generate_audio(text, voice, response_format, speed): 23 | # Determine if the voice is an OpenAI-compatible voice or a direct edge-tts voice 24 | edge_tts_voice = voice_mapping.get(voice, voice) # Use mapping if in OpenAI names, otherwise use as-is 25 | 26 | # Generate the TTS output in mp3 format first 27 | temp_output_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") 28 | communicator = edge_tts.Communicate(text, edge_tts_voice) 29 | await communicator.save(temp_output_file.name) 30 | 31 | # If the requested format is mp3 and speed is 1.0, return the generated file directly 32 | if response_format == "mp3" and speed == 1.0: 33 | return temp_output_file.name 34 | 35 | # Convert to the requested format if not mp3 or if speed adjustment is needed 36 | converted_output_file = tempfile.NamedTemporaryFile(delete=False, suffix=f".{response_format}") 37 | 38 | # ffmpeg playback speed adjustment 39 | speed_filter = f"atempo={speed}" if response_format != "pcm" else f"asetrate=44100*{speed},aresample=44100" 40 | ffmpeg_command = [ 41 | "ffmpeg", "-i", temp_output_file.name, 42 | "-filter:a", speed_filter, # Apply speed adjustment 43 | "-f", response_format, "-y", 44 | converted_output_file.name 45 | ] 46 | 47 | try: 48 | subprocess.run(ffmpeg_command, check=True) 49 | except subprocess.CalledProcessError as e: 50 | raise RuntimeError(f"Error in audio conversion: {e}") 51 | 52 | return converted_output_file.name 53 | 54 | def generate_speech(text, voice, response_format, speed=1.0): 55 | return asyncio.run(_generate_audio(text, voice, response_format, speed)) 56 | 57 | def get_models(): 58 | return [ 59 | {"id": "tts-1", "name": "Text-to-speech v1"}, 60 | {"id": "tts-1-hd", "name": "Text-to-speech v1 HD"} 61 | ] 62 | 63 | async def _get_voices(language=None): 64 | # List all voices, filter by language if specified 65 | all_voices = await edge_tts.list_voices() 66 | language = language or DEFAULT_LANGUAGE # Use default if no language specified 67 | filtered_voices = [ 68 | {"name": v['ShortName'], "gender": v['Gender'], "language": v['Locale']} 69 | for v in all_voices if language == 'all' or language is None or v['Locale'] == language 70 | ] 71 | return filtered_voices 72 | 73 | def get_voices(language=None): 74 | return asyncio.run(_get_voices(language)) 75 | -------------------------------------------------------------------------------- /src/api/app/server.py: -------------------------------------------------------------------------------- 1 | # server.py 2 | 3 | from flask import Flask, request, send_file, jsonify, render_template_string 4 | from gevent.pywsgi import WSGIServer 5 | from dotenv import load_dotenv 6 | import os 7 | 8 | from tts_handler import generate_speech, get_models, get_voices 9 | from utils import require_api_key, AUDIO_FORMAT_MIME_TYPES 10 | 11 | app = Flask(__name__) 12 | load_dotenv() 13 | 14 | API_KEY = os.getenv('API_KEY', 'your_api_key_here') 15 | PORT = int(os.getenv('PORT', 5050)) 16 | 17 | DEFAULT_VOICE = os.getenv('DEFAULT_VOICE', 'en-US-AndrewNeural') 18 | DEFAULT_RESPONSE_FORMAT = os.getenv('DEFAULT_RESPONSE_FORMAT', 'mp3') 19 | DEFAULT_SPEED = float(os.getenv('DEFAULT_SPEED', 1.0)) 20 | 21 | # DEFAULT_MODEL = os.getenv('DEFAULT_MODEL', 'tts-1') 22 | 23 | # 添加一个简单的 HTML 模板 24 | HOME_TEMPLATE = """ 25 | 26 | 27 | 28 | 29 | 30 | OpenAI通用的微软Edge TTS API 31 | 36 | 37 | 38 |

OpenAI通用的微软Edge TTS API

39 |

这是一个基于 Edge TTS 的 OpenAI TTS API 替代品。

40 |

API 端点：

41 | 47 |

使用示例：

48 | 
 49 |     

 50 |     视频教程
 51 |     更多信息请参考项目说明

 52 |
53 | 54 | 55 | """ 56 | 57 | @app.route('/') 58 | def home(): 59 | return render_template_string(HOME_TEMPLATE, port=PORT) 60 | 61 | @app.route('/v1/audio/speech', methods=['POST']) 62 | @require_api_key 63 | def text_to_speech(): 64 | data = request.json 65 | if not data or 'input' not in data: 66 | return jsonify({"error": "Missing 'input' in request body"}), 400 67 | 68 | text = data.get('input') 69 | # model = data.get('model', DEFAULT_MODEL) 70 | voice = data.get('voice', DEFAULT_VOICE) 71 | 72 | response_format = data.get('response_format', DEFAULT_RESPONSE_FORMAT) 73 | speed = float(data.get('speed', DEFAULT_SPEED)) 74 | 75 | mime_type = AUDIO_FORMAT_MIME_TYPES.get(response_format, "audio/mpeg") 76 | 77 | # Generate the audio file in the specified format with speed adjustment 78 | output_file_path = generate_speech(text, voice, response_format, speed) 79 | 80 | # Return the file with the correct MIME type 81 | return send_file(output_file_path, mimetype=mime_type, as_attachment=True, download_name=f"speech.{response_format}") 82 | 83 | @app.route('/v1/models', methods=['GET', 'POST']) 84 | @require_api_key 85 | def list_models(): 86 | return jsonify({"data": get_models()}) 87 | 88 | @app.route('/v1/voices', methods=['GET', 'POST']) 89 | @require_api_key 90 | def list_voices(): 91 | specific_language = None 92 | 93 | data = request.args if request.method == 'GET' else request.json 94 | if data and ('language' in data or 'locale' in data): 95 | specific_language = data.get('language') if 'language' in data else data.get('locale') 96 | 97 | return jsonify({"voices": get_voices(specific_language)}) 98 | 99 | @app.route('/v1/voices/all', methods=['GET', 'POST']) 100 | @require_api_key 101 | def list_all_voices(): 102 | return jsonify({"voices": get_voices('all')}) 103 | 104 | print(f" Edge TTS (Free Azure TTS) Replacement for OpenAI's TTS API") 105 | print(f" ") 106 | print(f" * Serving OpenAI Edge TTS") 107 | print(f" * Server running on http://localhost:{PORT}") 108 | print(f" * TTS Endpoint: http://localhost:{PORT}/v1/audio/speech") 109 | print(f" ") 110 | 111 | if __name__ == '__main__': 112 | http_server = WSGIServer(('0.0.0.0', PORT), app) 113 | http_server.serve_forever() 114 | --------------------------------------------------------------------------------