Features
Dictate anywhere.Keep everything.
One key in every app. Recognition on your Mac. A vocabulary that is yours, and a record you can search and measure.
What ships
System-wide dictation
Press the key, speak, release. The words land where your cursor is — a document, a chat, a terminal, a prompt. No app to switch to, no mode to remember.
On-device recognition
Recognition happens on your machine, with no cloud transcription path for it to leave through. Works with no network.
Two engines, one take
A fast engine writes words as you say them; an accurate engine upgrades the transcript from behind while you are still talking. Most of the work is done before you release the key.
Learns your words
The names, terms and products you use are harvested from your own writing and from what your agents write back — spelled once, correctly — then supplied to the engine as you speak.
Context while you talk
What you just said summons the terms most likely to come next, mid-take. The payload changes as the conversation moves, and every take reports what it was given and what landed.
Reads the screen you are in
Visible text in the app you are dictating into is matched against words you have used before — never free text, never sent anywhere.
Corrections train it
An edit you make to inserted text is offered back as a suggestion. Accept it and the spelling is yours from then on.
Studio
Everything you have said, searchable on your Mac. Your vocabulary, characterised rather than listed. How each engine does on your voice, measured on your own takes.
Chooses by measurement
Engine and model choices are made on your own recordings, not on a leaderboard. The benchmark runs on what you actually said.
Every take, kept
Each take is stored with its audio and every pass over it. A better engine later can be run over everything you have ever said — nothing is rewritten, only added beside.