Features
Push to talk
Hold a modifier key, speak, release. Text lands in the frontmost app.
On-device recognition
Runs on the Neural Engine. Works with no network at all.
File transcription
Audio and video in, text or subtitles out. No length limit.
SRT, VTT, Markdown
Export with real timings, taken from the decoder rather than guessed.
Custom vocabulary
Register the names any model gets wrong. They stop being wrong.
AI rewriting
Your own instruction, applied to every dictation. Paid.
Per-app routing
A Slack message and an email do not have to sound the same.
App blocklist
Mark apps where rewriting never runs, whatever else is set.
Never steals focus
The recording indicator cannot take key status, so insertion never misfires.