Skip to project

Local dictation · 2026

MattyFlow

Hold Fn to dictate into your current app, with local transcription and optional rewriting.

Native appFn Record audioRemember the destination
Local serviceTranscribeOptional rewrite
Native appInsert the resultNewest request only
Recording and insertion stay in the native app; Python runs the models.

Recording and insertion

Capture & return
The app remembers which application was focused when recording began. Only the newest completed request is pasted there.
Model service
FastAPI prepares the recording for the selected speech model and, when enabled, sends its transcript through a rewrite prompt.
Model lifecycle
Models download separately and can unload when idle. Loading them again adds time to the next request.

Models and permissions

Models run on the Mac by default. OpenAI is also available as an optional provider.

Requires Apple silicon and Microphone, Input Monitoring, and Accessibility permissions.