Utter

Utter turns speech into text in whichever app has the cursor. You hold fn, or another modifier key, speak, and let go; the transcript is typed in. The speech model runs on the Mac, with whisper.cpp or an optional engine, and audio is never uploaded. The app also keeps a searchable library of past dictations, can turn one into a note, offers to transcribe a call, and has a personal dictionary and text snippets. It suits people who dictate into many apps and want recognition that works offline and without a subscription.

Utter runs on macOS 14 or later, on Apple silicon or Intel. The DMG includes a speech model of about 190 MB. The README says the app is signed but not notarized, so the first launch needs Open Anyway under System Settings, Privacy & Security. It asks for Microphone and Accessibility permissions, and it has no accounts or telemetry.