Product Launch

WhisperDictate — Voice Dictation for Mac

3 min read

Download WhisperDictate

No account needed. Works on macOS 14 (Sonoma) or later.

Download for Mac (.dmg)

~2 MB download • Speech model (~142 MB) downloads on first launch

A native Mac app built with Swift — just drag it to Applications from the DMG and you're good.

Under the hood it uses WhisperKit, a Swift framework that runs OpenAI's Whisper speech-to-text model directly on your Mac using Apple's Neural Engine and Metal GPU. It's the same AI model behind Whisper, just optimized specifically for Apple silicon — so transcription is fast and runs 100% locally. No internet, no cloud, no API keys, no account, no subscription.

You press the Globe key → it records from your mic → transcribes locally in a couple seconds → auto-pastes wherever your cursor is. It also keeps a clipboard history of everything you've dictated so you can search and re-use past transcriptions.

How It Works

Hold Mode

Hold the Globe key (🌐 bottom-left of keyboard) → speak → release → text appears wherever your cursor is.

👆

Tap Mode

Quick-tap the Globe key → speak as long as you want → tap again → text appears. Great for longer dictation.

A small colored bar appears at the bottom of your screen to show status:

Red — Recording
Orange — Transcribing
Green — Done, text pasted

Install (~2 minutes)

  1. 1Open WhisperDictate.dmg and drag the app to Applications
  2. 2Right-click the app → Open (required once since it's not from the App Store)
  3. 3The speech model downloads automatically (~142 MB, takes about 20 seconds)
  4. 4Grant Accessibility permission when prompted — go to System Settings → Privacy & Security → Accessibility and toggle WhisperDictate on
  5. 5Grant Microphone access when prompted
  6. 6Done! Look for the waveform icon (⎍) in your menu bar

Tips

  • Click the waveform icon in the menu bar to see status and settings
  • Enable "Launch at Login" from the menu bar dropdown so it's always ready
  • If you see a ⚠️ instead of the waveform, Accessibility permission needs to be granted — click the icon for instructions
  • Works best on Apple Silicon Macs (M1/M2/M3/M4) — Intel works but transcription is slower

🔒 Privacy First

All transcription happens locally on your Mac using WhisperKit (Apple CoreML + Neural Engine). No audio or text is ever sent to any server. The speech model is downloaded once from Hugging Face on first launch and stored locally.