What one key does
Speak, and it lands at your cursor
Hold, speak, release. The corrected text goes into the app and field your cursor was in when you started.
⌘C ⌘CTranslate a selection instantly
Korean to English, English to Korean. See the original, the translation, the actual model, and tokens in one window.
⌃⇧COne key for work on selected text
Pick correct, structure, translate, case conversion, send to Notes, or ask a CLI from a grid.
⌃⇧SShow your screen while you ask
Capture a region and speak; the image and the text are pasted into your agent or chat, in that order.
Session recordingRecord meetings, transcribe later
Record up to 3 hours, then create minutes and a verbatim transcript and attach speakers.
⌃⇧VSaved phrases to your cursor
Find addresses, account numbers, or templates by name. They go in exactly as saved, without correction.
Your first 10 minutes
- Install
Download the DMG and move it to Applications.
- Permissions and model
Allow Microphone and Accessibility in the tutorial and download a model.
- First dictation
Hold ⌃⇧R, say one sentence, and release.
- Connect an LLM
Add the provider and key used for correction and translation.
On your Mac by default
- Speech recognition runs on your Mac with WhisperKit or MLX Audio.
- LLM correction defaults to None. Text is sent only to the provider you pick.
- Translation, screen region, and AWS Transcribe do nothing until you turn them on.