Speak

macOS menu bar dictation tool powered by Deepgram Nova-3. Press a hotkey, talk, and text gets typed into any app in real-time.

Features

• Lives in the menu bar — no dock icon, no window clutter
• Real-time speech-to-text via Deepgram Nova-3 streaming
• Text injection into any focused text field (clipboard + paste)
• Floating pill widget with waveform visualizer
• Custom vocabulary (keyterms) for names, brands, technical terms
• Configurable hotkey, language (EN/RU/Multilingual), microphone
• Auto-save settings

Setup

# Clone
git clone https://github.com/iota-uz/speak.git
cd speak

# Install
npm install

# Get a Deepgram API key at https://console.deepgram.com
# You'll enter it in the app settings

# Run
npm start

On first launch, macOS will ask for microphone and accessibility permissions. Grant both.

Usage

1. Click the tray icon to open settings
2. Enter your Deepgram API key
3. Set your preferred hotkey (default: Cmd+Shift+Z)
4. Select microphone and language
5. Press the hotkey to start dictating — a floating pill appears
6. Speak — text appears in real-time in whatever app is focused
7. Press the hotkey again to stop

Cost

Deepgram Nova-3 streaming: ~$0.36/hour. Realistic usage for a product manager is $3-7/month.

Requirements

• macOS (uses macOS-specific accessibility APIs for text injection)
• Node.js 18+
• Deepgram API key