YouTube Transcript API

A Python service that provides APIs to fetch and transcribe YouTube video content. It supports both REST API (Flask) and MCP server implementations.

Features

• Fetch YouTube video transcripts in multiple languages (English and Vietnamese)
• Auto-detect and use available transcripts
• Fallback to audio transcription using Whisper when transcripts are unavailable
• Support for both REST API and MCP server interfaces
• Automatic language detection
• Temporary file cleanup
• Progress reporting for long-running operations

Installation

pip install -r requirements.txt

Usage

REST API (Flask)

Start the Flask server:

python apps/flask_server.py

Available endpoints:

• GET /transcript?video_id=<video_id>&language=<lang> - Get video transcript
• GET /video/info?video_id=<video_id> - Get video information
• GET /health - Health check endpoint

MCP Server

Start the MCP server:

python apps/mcp_server.py

Available tools:

• get_transcript(video_id, language) - Get video transcript
• extract_transcript(video_id, language) - Extract transcript from audio
• search_youtube_video(query) - Search for YouTube videos

Language Support

• English (en)
• Vietnamese (vi)
• Auto-detection for other languages

Dependencies

• youtube-transcript-api
• pytube
• whisper
• torch
• langdetect
• flask (for REST API)
• mcp (for MCP server)

Development

The project structure:

apps/
├── __init__.py
├── flask_server.py  # REST API implementation
├── mcp_server.py    # MCP server implementation
└── utils.py         # Shared utilities

License

MIT License