More accurate summaries and transcripts. • Summaries drop placeholder items like "None", and action items keep their due dates • Interview and legal summaries add verbatim quotes, each checked word-for-word against the transcript, with the speaker and a tap-to-play time • Mixed Chinese–English meetings are summarised in the right language • Chinese transcripts follow your Simplified or Traditional choice • Phrases invented during silence are filtered out • "Check audio" hints now point to individual sentences • Renaming a speaker updates the summary too • Exports include concerns, quotes, keywords and due dates • Long recordings are summarised in full with the downloadable summary model • A reminder to confirm everyone has agreed before you record — turn it off in Settings if you only record yourself • On iPhone Duo, unfolding shows your recordings and the open transcript side by side

Fixes a crash on long recordings. Recordings over about 40 minutes could run out of memory during transcription and quit — most often on the hour-long interviews and sessions this app is built for. Audio is now transcribed in segments, so length no longer drives memory use. Also in this update: • Progress now advances throughout transcription instead of appearing stuck • Transcripts saved in earlier versions are read correctly after upgrading

• Speaker labeling rebuilt on a neural speaker-embedding model — far more accurate for two-person recordings • AI summaries now cover the whole recording, not just the opening minutes • Transcripts and summaries are now encrypted at rest, alongside your audio • Custom vocabulary now applies to the main transcription engine • Fixed Chinese full-text search • Free trial recordings extended from 20 to 60 minutes each • Download progress is now shown when models are fetched for the first time