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
New transcription engines: gpt-transcribe for final transcripts and gpt-live-transcribe for the live preview, selectable per path in Settings → Transcriptions.
New Credentials pane holds the OpenAI API key (stored in the Keychain) with a one-click test button. The OpenAI engine options only appear once a key is set.
Switch the live preview between on-device and OpenAI at any point during a meeting; the configured engine returns for the next meeting.
Live audio is upsampled to 24 kHz for the realtime API, and streamed transcripts arrive per sentence.
Live preview
The current sentence now grows in place and closes at punctuation, so the preview flows continuously with readable lines instead of chunk fragments.
Speaker labels are no longer shown before each preview line.
Meeting threads
All user-facing "Codex" wording is now "ChatGPT", and thread models are picked from ChatGPT's live model list again.
The summary model picker moved to the Summaries pane.
Settings
The sidebar groups ChatGPT, Tana, and Hooks under Integrations.
Ignored calendar events are managed as removable pills with an add field.
The Dictionary editor moved into the Transcriptions pane.
Assorted layout and copy cleanups throughout.
Fixes
Yesterday's calendar event title no longer lingers in the title box the next morning.
Grouped week-based archive folders and storage usage view (from the 1.3.0 betas).