Faster, More Reliable On-Device Transcription (v3.7.1)

On macOS and iOS, sessions using the Nemotron speech engine now finish processing much faster after you end them: when the live transcript completed cleanly, Hedy skips re-transcribing the entire recording and goes straight to speaker attribution and your recap. Long sessions may still take the thorough full pass when needed for accuracy.
On desktop, live audio frames are now guaranteed to reach the transcription engine in order, preventing rare transcript glitches during long sessions.