Support

Only the first time. That's when the speech model and the speaker model are downloaded, about 500–650 MB depending on the language, so Wi-Fi is a good idea. You can pause the download and continue later. After that everything works offline, even in Airplane Mode. If you add a language that uses a different model, that model is downloaded when you ask for it.

Swedish, English and 23 other European languages: Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish and Ukrainian. Swedish uses Klang Pianissimo, a model trained specially for Swedish. The other languages use NVIDIA Parakeet. Choose your first language, and an optional second one, in the app's settings. To record in the second language, touch and hold the bubble and drag diagonally up to the right. If a recording ended up in the wrong language, you can transcribe it again in another language from the share menu.

The speech model is loaded when the app starts. You can start talking right away anyway: the audio is saved from the first second and the text catches up when the model is ready.

Open Settings › Apps › Snicksnack and turn on Microphone. Without the microphone, the app can still transcribe audio files.

Yes. You can put the phone in your pocket. If someone calls, recording pauses and everything up to that point is saved. Tap Continue when the call is over.

While you talk, the app writes in short pieces to keep up. After stop, the whole recording is transcribed again in one pass with all the context, which gives better text. The words that changed light up briefly. The speakers are sorted out at the same time.

Put the phone close to whoever is talking, ideally in the middle of the table, and try not to talk over each other. If something comes out wrong, you can edit it right in the text, and playback still lands in the right place.

Open the recording and tap a speaker, for example “Speaker 2”, to give them a name. The name then changes everywhere in that recording. There is no fixed limit on how many speakers the app can tell apart.

Yes. In the app, add a receiver by scanning its QR code, then send a recording there from the Send menu. Nothing is sent on its own. You can also start a recording from Apple Watch. Read more under Privacy.

Share the file from another app and choose Snicksnack, or tap “…” at the top right of the list and choose to transcribe an audio file. It works with most audio and video formats, such as voice memos and voicemails.

In the app, and in the Files app under On My iPhone › Snicksnack. The audio is compressed without loss of quality and takes about 1 MB per minute. Recordings are not included in your iCloud Backup unless you turn that on in the Settings app under Snicksnack. To delete a recording, swipe left in the list, or touch and hold it.

Yes. When you transcribe an audio file or tap stop, Snicksnack keeps working in the background and the progress shows on the Lock Screen. The text appears in the recording as it goes, so you can start reading before everything is done.

iPhone 11 or later with iOS 26. Transcription and speakers work on all of them. Summaries need an iPhone with Apple Intelligence (iPhone 15 Pro or later) with Apple Intelligence turned on.