Cadence - Audio Transcription

Cadence turns an audio or video recording into a timestamped transcript that knows who was speaking. It transcribes on your iPhone or iPad: no account, no upload, and no per-minute billing. Import an interview, a podcast or a meeting recording from Files, and read it back as speaker-labelled turns instead of a wall of text. WHY CADENCE • Transcribes on device, so a recording you are not allowed to upload stays with you • Two engines are free and unlimited: no trial, no watermark, no sign-up, no cap • Speaker diarization included, not a separate paid add-on • Works offline once a model is downloaded KEY FEATURES Import from Files or any app • Pick one recording or several at once, from Files or iCloud Drive • Share a recording from Voice Memos or another app, and it can start transcribing on its own • Video is transcribed from its audio track, so screen and meeting recordings work directly • Cadence keeps its own copy, so a recording still plays after the original file moves Choose your engine • Apple Speech is built into iOS, and it is the free default • Parakeet TDT v3 compact covers 25 European languages, free; full v3, Parakeet Ultra and the English-tuned v2 are Pro • Whisper large-v3-turbo for 100 languages, detected automatically (Pro) • Download a model once, and delete it in Settings when you want the space back Know who said what • Speaker diarization via pyannote Community-1, a 22 MB download • Speaker-attributed turns with timecodes, in colors shared by legend and transcript • Rename a speaker once and every turn follows • Name someone once and Cadence offers to recognize their voice in other recordings (Pro) • Tell Cadence how many speakers to expect, for the recordings that need the help Play it back and read along • Play the source from 0.75x to 2x, with each word marked as it is said • Skip back or ahead 15 seconds, and let the transcript follow playback • Tap a turn to play from there Find anything you said • Search recording names and transcript text together, accents and case folded, so "jose" finds "José" • Open a result and its matches are highlighted in the transcript • Select and copy any passage • Export to plain text and SRT, free, and save it anywhere Files can reach More room on iPad • The library and the transcript side by side, in any orientation and in Split View • Drag recordings in from Files, Voice Memos or Mail • Keyboard shortcuts for playback and find CADENCE PRO Free covers the core on two engines: Apple Speech, and Parakeet TDT v3 compact for 25 European languages. Unlimited transcription, diarization, playback and search. One purchase, with no subscription, adds: • The full-size speech models: Parakeet TDT v3 at full precision, Parakeet Ultra, the English-tuned v2, and Whisper for 100 languages • A queue, so you can start several recordings and Cadence works through them in turn • Briefs: decisions, action items and open questions, or a summary of one turn, made on device • On iPad, two models compared side by side, with the differences marked • Export to WebVTT, Markdown, JSON and CSV • The speaker library, which recognizes the people you have named in other recordings Pro is one purchase for your iPhone, your iPad and your Mac. Cadence for Mac also adds Transcribe All. WHO IT'S FOR • Researchers and journalists working through interview recordings • Podcasters checking what was said in an episode • Anyone with meeting or lecture recordings, only useful once you can tell the speakers apart PRIVACY • Cadence never uploads your audio or your transcripts • No account, no analytics, and no third-party data collection • The only network traffic is downloading models once: the speech models you choose, and the 22 MB speaker model REQUIREMENTS iPhone or iPad with iOS 26 or later. Briefs need Apple Intelligence. Keep Cadence open while a long recording transcribes.