Video Extraction Server

What is Video Extraction Server?

Video Extraction Server is an MCP (Model Context Protocol) server that provides text extraction from video platforms and audio files. It runs as a Python-based server and is designed for users who need to transcribe audio from online videos or local audio files through any MCP-compatible client, such as Claude Desktop.

How to use Video Extraction Server?

Installation requires Python 3.9+, FFmpeg, and the Whisper model (downloaded automatically on first run). You can either clone the repository and run python server.py after installing dependencies, or install via pip install mcp-video-service and configure an MCP client with the command uvx mcp-video-service. The server exposes four tools: video download, audio download, video text extraction, and audio text extraction, each callable via MCP tool requests.

Key features of Video Extraction Server

• High-quality speech recognition using OpenAI Whisper
• Multi-language text recognition
• Support for audio formats including mp3, wav, m4a
• MCP-compliant tools interface for standardized access
• Asynchronous processing for large files
• Downloads from YouTube, Bilibili, TikTok, Instagram, Twitter/X, Facebook, Vimeo, Dailymotion, SoundCloud, and more

Use cases of Video Extraction Server

• Transcribe YouTube or Bilibili videos for content analysis
• Extract spoken text from TikTok or Instagram clips
• Convert local audio recordings (lectures, meetings) into searchable text
• Build automated transcription pipelines integrated with MCP-enabled applications

FAQ from Video Extraction Server

Which platforms does Video Extraction Server support?

It supports YouTube, Bilibili, TikTok, Instagram, Twitter/X, Facebook, Vimeo, Dailymotion, SoundCloud, and many others via yt-dlp. See yt-dlp supported sites for the full list.

What are the system requirements?

Minimum 8 GB RAM and FFmpeg are required. GPU acceleration (NVIDIA + CUDA) is recommended for faster processing. Sufficient disk space is needed for the Whisper model (≈1 GB) and temporary files.

How do I choose the Whisper model size?

Edit config.yaml to set the model to tiny, base, small, medium, or large. Smaller models are faster but less accurate; large offers highest accuracy but uses more resources.

What happens on the first run?

The server automatically downloads the Whisper model (≈1 GB) on first startup. This may take several minutes to tens of minutes depending on network speed. The model is cached locally for subsequent runs.

Can I use Video Extraction Server with any MCP client?

Yes. It works with any MCP-compatible client, including Claude Desktop and custom MCP applications. Configuration typically involves specifying the command (uvx mcp-video-service or pointing to the Python script) in the client’s MCP settings.