Skip to content
MCP server

Local Whisper

Transcribe audio and video with Whisper on your own machine, for Claude users

Runs locallyStatus and history
Install
Platforms
Windows, macOS, Linux
Runs
Locally
Version
1.1.0
Listing updated
today

Similar MCP servers

About Local Whisper

Local Whisper is an MCP server for design and media. It runs locally on your machine and is started by your AI client.

Everything runs on the local computer: the server hands audio to a Whisper model, so voice notes, meetings and lectures are not uploaded anywhere. It uses the GPU through Vulkan on Windows and Linux or Metal on Apple Silicon, and drops back to the CPU if no GPU is usable. The README reports about one second per minute of audio on an RTX 3080 with the default large-v3-turbo-q8_0 model.

It reads WAV, Ogg Opus (the format of WhatsApp and Telegram voice notes) and MP3 out of the box. With ffmpeg on the PATH it also handles M4A and FLAC, and it can pull the audio track out of MP4 or WebM videos. Output can be plain text, SRT subtitles or JSON segments with timestamps.

The model is downloaded on first use and checked against the hash Hugging Face publishes, then unloaded after five minutes idle to free memory. It ships as one native executable per platform, either as a Claude Desktop extension (.mcpb) or a binary registered with Claude Code, and an optional skill tells Claude when to call it.

Install

Run in a terminal

claude mcp add local-whisper-mcp -- npx -y https://github.com/OrihuelaConde/local-whisper-mcp/releases/download/v1.1.0/local-whisper-mcp-1.1.0-win-x64.mcpb

Add to claude_desktop_config.json

{
  "mcpServers": {
    "local-whisper-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/OrihuelaConde/local-whisper-mcp/releases/download/v1.1.0/local-whisper-mcp-1.1.0-win-x64.mcpb"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "local-whisper-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/OrihuelaConde/local-whisper-mcp/releases/download/v1.1.0/local-whisper-mcp-1.1.0-win-x64.mcpb"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "local-whisper-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/OrihuelaConde/local-whisper-mcp/releases/download/v1.1.0/local-whisper-mcp-1.1.0-win-x64.mcpb"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "local-whisper-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/OrihuelaConde/local-whisper-mcp/releases/download/v1.1.0/local-whisper-mcp-1.1.0-win-x64.mcpb"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "local-whisper-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/OrihuelaConde/local-whisper-mcp/releases/download/v1.1.0/local-whisper-mcp-1.1.0-win-x64.mcpb"
      ]
    }
  }
}

Using several servers? Build one config file with the MCP config generator.

Tools this server provides (3)

Taken from the project documentation; the running server may differ.

transcribe
Transcribes an audio or video file; takes path, model, language, format (txt, srt, json), vad and hints.
status
Reports the device in use, installed models and sizes, downloads in progress, inbox, allowed folders and log folder.
delete_models
Deletes downloaded models to free disk space.

What it does

Fully local transcription
Audio never leaves the computer; Whisper runs on your own hardware.
GPU with CPU fallback
Uses Vulkan or Metal acceleration and switches to the CPU on its own.
Audio and video input
Handles WAV, Ogg Opus, MP3, and with ffmpeg M4A, FLAC, MP4 and WebM.
Text, SRT or JSON output
Choose plain text, subtitles, or timestamped JSON segments.
Voice activity detection
Can skip silence on clean recordings or transcribe everything on noisy ones.
Verified model download
Fetches the model once and checks it against Hugging Face's published hash.
Self-contained binaries
One native executable per platform with no Python or .NET runtime.

Good for

  • Turn a WhatsApp or Telegram voice note into text inside a Claude chat
  • Summarize decisions from a recorded meeting
  • Generate SRT subtitles for a class video
  • Transcribe lectures without sending them to a cloud service

Local Whisper FAQ

Which platforms does it support?

Windows x64, macOS on Apple Silicon and Intel, and Linux x64 and arm64.

Do I need ffmpeg?

Not for WAV, Ogg Opus or MP3. ffmpeg is needed for M4A, FLAC and video files.

Which Claude apps can use it?

Claude Desktop, Claude Code, and claude.ai chats that run on your computer through the desktop app.

What tools does it expose?

transcribe, status and delete_models.

How do I add Local Whisper to Claude Code?

Run this command in your terminal: claude mcp add local-whisper-mcp -- npx -y https://github.com/OrihuelaConde/local-whisper-mcp/releases/download/v1.1.0/local-whisper-mcp-1.1.0-win-x64.mcpb

What are the best alternatives to Local Whisper?

The closest alternatives to Local Whisper are HyperFrames by HeyGen, headroom, omniget, next-ai-draw-io and Ha MCP.

Where this comes from

  • Description: written from the project's README.
  • Install steps: from the official MCP Registry entry.

How AIPoweredBot checks servers