*Press "Control + ~" → Speak → Text appears instantly at your cursor*

🚀 Lightning-fast AI transcription directly at your cursor - The fastest way from voice to text

MLX-Transcribe is a NATIVE SIGNED privacy-first, lightning-fast transcription app for macOS that transforms your voice into text instantly at your cursor position. Built with Apple Silicon optimization and seamless integration with MLX-GUI, it's the perfect tool for developers, writers, and anyone who values speed and privacy.

• ⚡ Instant Transcription - Press Control + ` and speak - text appears instantly at your cursor
• 🤖 Local AI First - Seamless MLX-GUI integration with automatic model management
• 🛡️ Privacy Focused - Your voice never leaves your machine with local processing
• 🎛️ Multi-API Support - Works with Local MLX, OpenAI, and Groq endpoints
• ⌨️ Customizable Hotkeys - Configure any key combination that works for you
• 🎯 Universal Cursor - Works in any app - text editors, browsers, chat apps, terminals
• 🚀 Menu Bar Native - Lightweight, always-accessible menu bar app
• 🔄 Auto-Install Models - Automatically downloads and loads Parakeet models via MLX-GUI
• 🏃‍♂️ Startup Ready - Optional login item for seamless workflow integration

1. Download the latest release from Releases
2. Open the .zip file and drag MLX-Transcribe to Applications
3. Launch and grant accessibility permissions when prompted

MLX-Transcribe will automatically request microphone access on first launch.

For text insertion at cursor:

1. Click "Grant Accessibility" in the menu bar popup
2. Add MLX-Transcribe to Privacy & Security → Accessibility
3. Restart the app after granting permission

• Local MLX (Default): Integrates with MLX-GUI - models auto-install
• OpenAI: Add your API key in settings
• Groq: Add your Groq API key for fast cloud transcription

1. Press Control + ~ (or your custom hotkey)
2. Speak clearly while you see the red recording indicator
3. Press Control + ~ to stop recording, or Control + ~ again
4. Text appears instantly at your cursor position

Click the 🎤 icon in your menu bar to:

• View transcription status
• Configure hotkeys and API settings
• Enable "Start on Login"
• Check permissions and connection status

• Best Results: Speak clearly and pause briefly between sentences
• Quick Stop: Press Control + ~ while recording to stop immediately
• Works Everywhere: Terminal, VS Code, Slack, browsers, documents
• Privacy Mode: Use local MLX processing to keep everything offline

Endpoint: http://127.0.0.1:8000/v1/audio/transcriptions
Model: parakeet-tdt-0-6b-v2 (auto-installed)
API Key: Not required

Endpoint: https://api.openai.com/v1/audio/transcriptions
Model: whisper-1
API Key: Required (sk-...)

Endpoint: https://api.groq.com/openai/v1/audio/transcriptions
Model: whisper-large-v3-turbo
API Key: Required (gsk_...)

• macOS: 12.0 (Monterey) or later
• Architecture: Intel x86_64 or Apple Silicon (arm64)
• Memory: 4GB RAM minimum, 8GB recommended
• Storage: 50MB for app, additional space for local models
• For Local MLX: MLX-GUI installed and running

Popular combinations:

• Control + ~ (Default) - Great for developers
• Shift + Tab - Alternative option
• Option + Space - Quick access
• F5 - Function key option

• Start/Stop Recording: Your configured hotkey (default: Control + ~)
• Force Stop: Control + ~ (while recording)
• Open Menu: Click menu bar icon

We welcome contributions! Here's how you can help:

1. 🐛 Report Issues: Found a bug? Open an issue
2. 💡 Feature Requests: Have an idea? We'd love to hear it!
3. 🔧 Code Contributions: Fork, branch, commit, and PR
4. 📖 Documentation: Help improve our docs and examples
5. ⭐ Star the Repo: Show your support!

• MLX-GUI - The powerful MLX interface that powers local transcription
• matthewrogers.org - Creator's website and blog

Freeware License - see LICENSE file for details.

• Built with ❤️ by Matthew Rogers
• Powered by MLX-GUI for local AI processing
• Uses Apple's Speech Recognition framework
• Inspired by the need for privacy-first transcription tools

Made with 🎤 and ⚡ for the developer community

MLX-Transcribe - Where voice meets code at the speed of thought