mcp-whisper-transcription

galacoder/mcp-whisper-transcription

3.2

If you are the rightful owner of mcp-whisper-transcription and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

The MCP Whisper Transcription Server is an advanced transcription server optimized for Apple Silicon devices, utilizing MLX-optimized Whisper models for efficient audio and video transcription.

Tools

  1. transcribe_file

    Transcribe a single audio/video file

  2. batch_transcribe

    Process multiple files in a directory

  3. list_models

    Show available Whisper models

  4. get_model_info

    Get details about a specific model

  5. clear_cache

    Clear model cache

  6. estimate_processing_time

    Estimate transcription time

  7. validate_media_file

    Check file compatibility

  8. get_supported_formats

    List supported input/output formats

MCP Whisper Transcription Server

An MCP (Model Context Protocol) server for audio/video transcription using MLX-optimized Whisper models. Optimized for Apple Silicon devices with ultra-fast performance.

Features

  • 🚀 MLX-Optimized: Leverages Apple Silicon for blazing-fast transcription
  • 🎯 Multiple Formats: Supports txt, md, srt, and json output formats
  • 🎬 Video Support: Automatically extracts audio from video files
  • 📦 Batch Processing: Process multiple files in parallel
  • 🔧 MCP Integration: Full MCP protocol support with tools and resources

Installation

  1. Clone the repository:

    git clone https://github.com/galacoder/mcp-whisper-transcription.git
cd mcp-whisper-transcription

  2. Install Poetry (if not already installed):

    curl -sSL https://install.python-poetry.org | python3 -

  3. Install dependencies:

    poetry install

  4. Copy environment configuration:

    cp .env.example .env

Configuration

Edit .env to configure:

  • DEFAULT_MODEL: Choose from tiny, base, small, medium, large-v3, or large-v3-turbo
  • OUTPUT_FORMATS: Comma-separated list of output formats
  • MAX_WORKERS: Number of parallel workers for batch processing
  • TEMP_DIR: Directory for temporary files

Usage

As MCP Server

Add to your Claude Code configuration:

{
  "mcpServers": {
    "whisper-transcription": {
      "command": "poetry",
      "args": ["run", "python", "-m", "src.whisper_mcp_server"],
      "cwd": "/path/to/mcp-whisper-transcription"
    }
  }
}

Available MCP Tools

  • transcribe_file: Transcribe a single audio/video file
  • batch_transcribe: Process multiple files in a directory
  • list_models: Show available Whisper models
  • get_model_info: Get details about a specific model
  • clear_cache: Clear model cache
  • estimate_processing_time: Estimate transcription time
  • validate_media_file: Check file compatibility
  • get_supported_formats: List supported input/output formats

Available MCP Resources

  • transcription://history - Recent transcriptions
  • transcription://history/{id} - Specific transcription details
  • transcription://models - Available models
  • transcription://config - Current configuration
  • transcription://formats - Supported formats
  • transcription://performance - Performance statistics

Development

Running Tests

poetry run pytest

Code Formatting

poetry run black .
poetry run isort .

Type Checking

poetry run mypy src/

Requirements

  • Python 3.9+
  • Apple Silicon Mac (for MLX optimization)
  • ffmpeg (for video file support)

License

MIT License - see LICENSE file for details

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Acknowledgments