Whisper Menu Bar

A minimal, clean speech-to-text menu bar application for macOS using OpenAI's Whisper. You can download the script here.

Features

• 🎤 Push-to-talk: Hold Option key to record, release to transcribe
• 📋 Auto-clipboard: Transcribed text automatically copied to clipboard
• 🔄 Model selection: Switch between tiny, base, small, and medium models
• 🎯 Clean & minimal: Simple interface, ~300 lines of code

Requirements

• macOS (tested on macOS 10.15+)
• Python 3.8 or higher
• Microphone access permissions

Installation and Usage

1. Copy the whisper-push-to-talk.py somewhere

2. Install uv and run:
   

   uv run whisper-push-to-talk.py

Note: On macOS, you may need to install PortAudio first for PyAudio:

   brew install portaudio

1. Grant microphone permissions to Terminal/your Python app when prompted

The app will:

1. Show a microphone icon (🎤) in your menu bar
2. Load the Whisper model in the background (first run may take a moment)
3. Display "Ready" when ready to use

Recording

• Option key: Hold to record, release to transcribe, Text appears in clipboard → Ready to paste anywhere
• Click the menu bar icon → Model → Select your preferred model:
  • tiny: Fastest, lowest accuracy (~1GB)
  • base: Good balance (default, ~1GB)
  • small: Better accuracy (~2GB)
  • medium: Best accuracy (~5GB)

Quit

Click the menu bar icon → Quit

License

MIT License - Feel free to modify and distribute

Credits

Built with OpenAI's Whisper model for speech recognition.