Voice to Text: Transcribe

Speech is recognised on the iPhone itself — no internet, no account. Meetings, lectures and interviews become text, split by speaker.

Record a meeting or add an audio file you already have — the app turns speech into text, splits it by speaker, and builds a summary with decisions and action items.

The important part happens on the iPhone itself. Recordings, transcripts, speakers and summaries stay in the phone's storage: no account, and no internet either, apart from one download of the speech models before the first run.

WHAT IT DOES • Recording with a timer, pause and resume; keeps going with the screen off • Import from Files or through the share sheet — for example from a voice memo app • On-device transcription: English, Russian and other languages, three quality modes • Speaker separation; names and roles are editable, and two speakers can be merged • Sectioned summary: what was discussed, decisions, action items, questions, risks, follow-ups, participants • Player with a waveform, scrubbing and 0.75×–2× speed: the transcript follows playback, and tapping a line seeks the audio • Search across meetings and inside a transcript • Export as TXT, Markdown, JSON or the original audio through the standard share sheet • Processing continues in the background and a notification tells you when it is ready • English and Russian interface, light and dark theme

PRIVACY • No account, no sign-up • No ads and no advertising identifiers • Usage statistics can be disabled in Settings and never include recordings, transcripts or meeting titles • Deleting a meeting removes its audio, transcript and summary

IF YOU WANT CLOUD PROCESSING Settings let you connect your own cloud speech key, and the transcript is then produced on that service. Until you do, the app shows no cloud screens at all, and before the first upload it asks separately, because the audio leaves the device. The key is stored in the iPhone Keychain only.

WHO IT IS FOR People who sit through calls and in-person meetings and would rather not retype them: managers, analysts, recruiters, researchers, journalists, students — and anyone who wants the conversation to stay off someone else's server.

GOOD TO KNOW Before the first transcription the app asks permission and downloads a model, between 150 MB and 1 GB depending on the quality mode. After that no internet is needed. Accuracy depends on the audio: the closer the phone is to the speakers, the better. Speaker separation is approximate — it can confuse similar voices and people talking over each other, so review the roles. On iPhones with Apple Intelligence the summary is written by the built-in model; on other devices a simpler key-phrase summary is generated.