Tap ⌘ twice
A small recorder appears at your cursor. No window stealing focus, no menu bar dance. Just a quiet indicator that the mic is hot.
Voice typing for macOS that listens locally, polishes on-device, and inserts at your cursor in any app. Built for terminals, IDEs, email, docs — anywhere you type.
⌘ twiceA small recorder appears at your cursor. No window stealing focus, no menu bar dance. Just a quiet indicator that the mic is hot.
Whisper turbo runs entirely on your Mac. Background noise, accents, technical jargon, code identifiers — handled. Latency lives below human reaction time.
⌘ twice againYour dictation lands at the cursor — fully punctuated, capitalized, and (if Pro) polished by a local LLM. Then it's gone. Nothing to clean up.
iTerm, Warp, Xcode, VS Code, Slack, Mail, Notion, Figma comments, password fields, Spotlight. Insertion uses Accessibility, not pasteboard hacks.
WhisperKit and a local Qwen 3.5 4B do the work. Cloud Polish (Pro, opt-in) sends post-transcription text only. We design for not having access.
The app reads which app is frontmost. In Terminal it preserves command names. In Mail it warms the tone. In Xcode it keeps camelCase. In Slack it relaxes.
Dictate the gist. Get the polished version. Math is resolved (10 × 2 → 20), filler is dropped, prose is restructured for clarity. Optional, transparent, instant.
Clean. Professional. Casual. Concise. Email. Bullet points. Genius. Switchable mid-flight via the menu bar — no preference panel acrobatics.
Funnel events are off until you opt in. Nothing about your dictation content is ever sent. Default outbound calls are model downloads from Hugging Face.
14-day money-back
Most popular
Local-only by design
Captured to a temp file, transcribed by WhisperKit, deleted. Never leaves the device. We don't have an audio endpoint to receive it.
Not collected, ever
Explicit non-collection clause covering BIPA (Illinois), CUBI (Texas), MyHealthMyData (WA), POST (NYC), CCPA-CPRA SPI.
Off by default
Opt-in funnel events (event names + plan/channel only). Never transcript content. Disable any time in preferences.
Text only · Pro · opt-in
Sends post-transcription text (no audio) to our authenticated proxy → OpenAI for the polish step you requested. Switch off any time.
requiresOnDeviceRecognition=true) transcribes locally. The temp audio file is deleted after each dictation. We don't operate a voice-receiving endpoint.