Support

No. There is no registration, no email address and no password — not even for the cloud. The App Store knows who paid; the app asks it.

Yes, completely. Apple's speech recognition needs no connection, and a downloaded Whisper model needs one even less. Only the cloud requires a connection.

Apple's speech recognition is instantly there and transcribes as you speak. Whisper understands considerably more, especially with dialect, but needs storage and loading time. In the Lab you can run several models over the same recording and see the difference yourself.

A Whisper model is prepared for the Neural Engine the first time — one to three minutes, once per model. After that it loads in seconds at app launch. Settings → On this device lets you get it out of the way with “Prepare”.

Seven days free with a limited allowance, then according to your plan. Billing is by seconds of speech, and you are charged only once the transcript is actually back — a cancelled or failed request costs nothing.

It passes through memory and is gone after the reply. Nothing is written to a disk, the transcript is stored nowhere, and the log holds a timestamp, a duration and a status code — no text, no file name, no IP address.

Yes. Under Settings → Server you enter the address and token of your own Whisper machine. The recording then goes there instead of to our cloud.

The system handles the transfer and catches up as soon as the iPhone is reachable again. Until then it waits on the watch and is listed under the record button. “Send now” nudges it along.

Delete individual notes in the list — the recording goes with them. Remove models under Settings → On this device. Deleting the app deletes everything: recordings, notes, settings and the token in the keychain.

Speech recognition is independent of the interface language. Apple's recognition covers the languages iOS supports on your device; Whisper understands around a hundred and copes best with dialect.