Free & open source for macOS
Hold a hotkey, speak, release. Transcription happens on-device in real time. No cloud. No subscriptions. No data leaves your Mac. Ever.
No servers. No latency. No privacy concerns. Just you and your Mac.
Hold the hotkey to record, release to transcribe. Natural as a walkie-talkie.
Text lands directly in whatever app you're using. No extra steps.
Every model runs on Apple Silicon. Zero network calls. Your audio is never uploaded.
Past transcriptions saved locally. Browse and copy anytime from the menu bar.
Lives in your menu bar. No dock icon, no window clutter. Always one hotkey away.
WhisperKit, Parakeet, Qwen3-ASR. Pick the model that fits your workflow.
No signup. No setup wizard. Install and go.
Download the .dmg, drag to Applications. The speech model downloads on first launch.
Press ⌥ + Space anywhere. A floating capsule appears while you talk.
Your words are transcribed on-device and pasted into the active app instantly.
State-of-the-art speech recognition models, all running locally on Apple Silicon.
Apple-optimized OpenAI Whisper. Multiple sizes from base to large-v3 turbo. Best overall accuracy and language coverage.
NVIDIA's 0.6B parameter model compiled for CoreML. Fast, lightweight, strong English transcription.
Alibaba's speech model on Apple MLX. Excellent with diverse accents and speaking styles.
Ready?
Free forever. No account required.