Zero cloud round-trips
Your audio is held in bounded memory while you dictate, then cleared. Transcript history and diagnostics are off by default.
Private, on-device dictation for Mac. Uttra turns your voice into polished text right where you’re working—without sending it anywhere.
Thanks again for today. I’ve attached the revised scope and moved the launch review to Friday.


Uttra’s privacy story is its architecture, not a setting buried in a menu. Capture, transcription, cleanup, and insertion all happen locally.
Your audio is held in bounded memory while you dictate, then cleared. Transcript history and diagnostics are off by default.
The speech model ships with the signed app. You don’t download one during onboarding.
The shipping transcription path refuses model downloads and retry-by-redownload behavior.
No account, no content logs, no advertising SDKs. Optional local diagnostics never include audio or transcript text.
Hold Right Option or tap Option–Space. Uttra listens from a compact island tucked beneath your Mac’s notch.
A bundled, reviewed speech model turns audio into text on your Mac. No upload, account, or first-run model download.
Uttra cleans up the result and inserts it at your cursor, so your thought lands where you were already working.
Correct a name or specialist term once. Uttra can remember your preferred spelling locally, so the words that matter keep coming out right.
Let Uttra detect the language automatically, or lock one in when you want complete control.
Uttra is coming to the Mac App Store for Apple silicon Macs. No account, no cloud processing, and no first-run model download.