You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Faster post-meeting refinement: each audio track is now prepared (voice detection, diarization, and embedding) in parallel, roughly halving preparation time for dual-track meetings with identical output.
Faster pre-recording start-up: the models required before recording now load in parallel, shortening the "Preparing" state.
Naming or marking a speaker during a meeting now registers them in Settings' speaker recognition in real time, so the same person is recognized across later meetings.
Fewer fragmented sentences in refined transcripts: same-speaker segments separated by short pauses are merged more sensibly, avoiding over-splitting long sentences into many small chunks.
Fixes
Renaming a speaker in the participant panel now updates already-rendered live subtitles, not just subsequent ones.
Fixed refined subtitles occasionally repeating a single character mid-sentence (an ASR decoder loop on long internal silence); refinement output now runs through the same cleaning as live captions, and repeated runs are collapsed anywhere in the text.
The "Overlapping speech" label now localizes in every supported language instead of always showing Chinese.
Downloads
macOS (Apple Silicon):Brevia-1.0.8-arm64.dmg, signed with Developer ID and notarized by Apple.