Whisper Transcription for Raycast

A Raycast extension that uses OpenAI's Whisper model for speech-to-text transcription, powered by whisper.cpp by Georgi Gerganov.

Features

• Start/stop audio recording with customizable hotkeys (default: Cmd+T)
• High-performance speech recognition using whisper.cpp
• Optimized for Apple Silicon
• Supports various audio formats via FFmpeg

Prerequisites

• Raycast
• FFmpeg
• whisper.cpp

Installation

1. Clone this repository:

   git clone https://github.com/yourusername/raycast-whisper-transcription.git
   

2. Install FFmpeg:

   brew install ffmpeg
   

3. Set up whisper.cpp:

   git clone https://github.com/ggerganov/whisper.cpp.git
   cd whisper.cpp
   make
   bash ./models/download-ggml-model.sh base.en
   

4. Add the extension to Raycast:

   • Open Raycast
   • Go to Extensions > Script Commands
   • Click the "+" button and select "Add Script Directory"
   • Choose the directory containing this extension

5. Set up the hotkey:

   • In Raycast, go to Extensions > Script Commands
   • Find the Whisper Transcription command
   • Click on it and assign your desired hotkey (e.g., Cmd+T)

Environment Setup

After installation, you need to set up the environment variables. Copy the .env.example file to a new file named .env and fill in the following variables:

1. Copy the example file:

   cp .env.example .env
   

2. Edit the .env file and set the following variables:

   RAYCAST_SCRIPT_DIR=/path/to/your/raycast/extension/directory
   WHISPER_DIR=/path/to/whisper.cpp/main
   WHISPER_MODEL=/path/to/whisper.cpp/models/ggml-base.en.bin
   

   Replace the paths with the actual locations on your system:

   • RAYCAST_SCRIPT_DIR: The directory where this Raycast extension is located
   • WHISPER_DIR: The directory containing the whisper.cpp main executable (typically /path/to/whisper.cpp/main)
   • WHISPER_MODEL: The full path to the Whisper model file (typically /path/to/whisper.cpp/models/ggml-base.en.bin)

Ensure that these paths are correct for your system setup. The extension will use these variables to locate the necessary files and directories.

Usage

1. Press Cmd+T (or your custom hotkey) to start recording audio
2. Press Cmd+T again to stop recording and start transcription
3. The transcribed text will appear in Raycast

How It Works

This extension leverages the high-performance C/C++ implementation of OpenAI's Whisper model provided by whisper.cpp. Key features include:

• Plain C/C++ implementation without dependencies
• Optimized for Apple Silicon via ARM NEON, Accelerate framework, Metal, and Core ML
• Mixed F16 / F32 precision
• 4-bit and 5-bit integer quantization support
• Zero memory allocations at runtime
• CPU and GPU inference support

License

MIT

Acknowledgements

• whisper.cpp by Georgi Gerganov
• OpenAI Whisper
• Raycast
• FFmpeg