Focus your summaries, fix your transcripts, and transcribe long recordings in full. • Summary range: choose where a summary starts and ends, so small talk before or after the part that matters is left out. The full transcript is kept. • Edit transcripts: correct misheard words passage by passage, or fix a name everywhere at once with Replace. Timings are kept, so tap-to-jump still lines up. • Long recordings now transcribe in full, in sections, instead of stopping short of the end. • Import audio recorded elsewhere and transcribe it — M4A, WAV, MP3, AIFF and CAF. • Choose the spoken language for transcription. A wrong automatic guess used to produce fluent text in the wrong language. • Each transcript now says which engine, model and language produced it, and whether it has been edited. • Read summaries aloud, in a natural voice, entirely on device. • Playback controls now stay on screen while audio plays, on both the transcript and summary screens. • Share, Edit and Re-transcribe sit at the top of the transcript.

Long recordings now transcribe in full, and your chosen microphone stays chosen. • Recordings over ten minutes are transcribed in sections. A 39-minute recording previously stopped short of the end; it no longer does. • The microphone you pick before recording is now held for the whole recording. A Bluetooth headset connecting mid-recording can no longer take the input away from it. • Recording recovers properly after a phone call. It previously gave up while the call still held the audio session. • Import audio recorded elsewhere and transcribe it — M4A, WAV, MP3, AIFF and CAF. • Choose the spoken language for transcription instead of leaving it to be detected. A wrong guess used to produce fluent text in the wrong language rather than an obvious error. • Each transcript now says which engine, model and language produced it. • Share and Re-transcribe moved to the top of the transcript. • Corrected the download sizes shown for AI summary models; several were understated by half. • Added an Acknowledgements page with the open-source licences the app is built on.

Clarion Transcribe is now free to download, with a 3-day free trial. • First-run setup: pick what you record most — meetings, interviews, lectures, medical or legal notes — and the app configures transcription and summaries to match. • Action items: meeting and interview summaries now pull out follow-ups, and anything with a date can be added to your calendar in a tap. • Much more accurate summaries. Names, numbers and figures are checked against the recording itself, so details that were never said don't end up in your summary. • Bullet summaries now actually come out as bullets. • Added the full Whisper Large v3 model for the most accurate transcripts. The app also picks a sensible model based on your device and available storage. • Manage downloaded speech and summary models in Settings to reclaim space. • Transcriptions queue instead of being dropped when one is already running. Everything still runs entirely on your device. Nothing is uploaded, and there is no account.

A big update to how Clarion Transcribe looks — and how you review your recordings. ALL-NEW DESIGN • A refreshed look drawn from our icon: deep ink blue with signal-blue and teal accents throughout. • A live waveform on the record screen that moves with your actual microphone level while you record. NAVIGATE YOUR RECORDINGS • Scrub anywhere in a recording with the new playback slider. • Skip back or ahead 10 seconds with one tap. • Transcripts are now timestamped passages — tap any passage to jump the audio straight to that moment, and follow along as the current passage highlights during playback. Tip: recordings transcribed before this update show plain text. Tap Re-transcribe on a recording to get the new tappable, timestamped transcript. As always: recording, transcription, and AI summaries all run entirely on your device. Nothing leaves your phone.

Transcription fixes and speed-ups: • Complete transcripts from Apple Speech on long recordings • Whisper no longer skips earlier audio or shows [BLANK_AUDIO] • Spanish and other languages now auto-detected by multilingual models • Large v3 Turbo model is now ~630 MB and much faster