• Find Recordings: bring in recordings you already have. Tap FIND on Sessions, or long-press the HarkNote icon and choose Import Recordings. Pick as many audio files as you like from Files or iCloud Drive, or share them from Voice Memos. They transcribe on your device like any other session, and nothing is uploaded. • Sharing several recordings to HarkNote at once now imports every one of them, not just the first.

• Naming other people's voices from saved profiles is now off by default. Turn it on under Config, Speaker Detection, Name known voices; a short note explains what it does first. Your own Self voice is matched either way.

• Auto-export: choose a folder under Config and every finished transcript and summary is written there as a text file, on its own. Same content as Share, without the share sheet. Off by default. • Every card in Config now collapses, with its current setting in the subtitle.

• Transcribe a finished recording again with a different model from the session menu, as long as the audio is still on your iPhone. • Summaries say plainly when no summarizer model is available. • Conversations transcribed with Parakeet now separate into speaker turns, even when people answer each other without pausing.

• Speaker labels now finish for hour-plus meetings, and a recording interrupted mid-way picks up where it left off instead of starting over. • The speaker-labels card shows elapsed time, an estimate, and which section it is working on.

• Apple Watch: record on your wrist and hand the audio to your iPhone for on-device transcription, with an optional live preview while you record. • Choose which microphone records, and refuse one you don't want. • The interface now ships in English, German, Spanish, French, Italian, Dutch, and Portuguese. • Changing the theme keeps your place in Settings; search dismisses the keyboard on return. • A recording with no speech says so instead of guessing. • Privacy policy v7 describes both transcription engines, Apple Foundation Models, and the Watch companion.

Choose your transcription engine. NVIDIA Parakeet v3 joins Whisper as a selectable on-device engine — still entirely on your iPhone. Take charge of your models. Download, switch between, and remove transcription models in Config → Transcription, with live download progress and a free-space check before anything is fetched. Plus: the Apple Watch companion records on your wrist even with no iPhone nearby, Ask is now a real multi-turn conversation per session, and the new Snow theme.

Redesigned recording, sessions, meetings, settings, onboarding, and privacy screens for a clearer, more consistent experience. Long recordings now process in resumable on-device sections with safer recovery, crash-safe transcript publication, automatic summaries, clearer progress and save receipts, and stronger original-audio protection. Also includes accessibility improvements and reliability fixes across recording, import, retry, and deletion.

Improved recording and transcription reliability, with smaller fixes and polish throughout the app.