Whisper MCP Server

Example Use Cases

• 1

  Transcribe single audio files on demand into text or subtitle formats (VTT, SRT, JSON).

• 2

  Batch transcribe large numbers of short audio files efficiently using GPU-accelerated batch processing and automatic batch-size tuning.

• 3

  A local MCP service integrated into desktop or automation workflows (e.g., Claude Desktop) to provide low-latency, on-premise speech recognition.

Description

This MCP server wraps Faster Whisper to offer fast, accurate speech-to-text transcription with support for multiple model sizes, automatic CUDA acceleration, and dynamic batch sizing. It supports batch transcription of folders, model instance caching to avoid reloads, and outputs in VTT, SRT, and JSON formats. The server includes audio validation and preprocessing, and provides tools to get model info, transcribe individual files, or batch transcribe folders. It's intended for easy integration into agent platforms (e.g., Claude Desktop) or automated pipelines requiring scalable transcription.