Privacy

Last updated November 2, 2026

Audio and text

The microphone is used only while you are recording. The audio is saved as a file in the app together with the text, the words and the speakers. Everything is transcribed by speech models that run on the phone. None of it is sent to us, and nothing is sent to anyone else unless you choose to (see Receivers below).

The same goes for audio files you choose to transcribe: they are copied into the app and transcribed on the phone.

What goes over the network on its own

The first time you open the app, a speech model (about 500–650 MB depending on the language) and a model that tells speakers apart (just over 20 MB) are downloaded from snicksnack-r2.applicerad.ai. If you later choose a language that is transcribed by a different model, that model is downloaded from the same place when you ask for it. The same happens if a model is updated. The download goes through Cloudflare, which, like any web server, sees your IP address and which files are downloaded. Nothing about you or your recordings is sent along, and we do not use this information to identify anyone. That is all the app does over the network on its own.

Where your recordings are

• In the app, and in the Files app under On My iPhone › Snicksnack.
• Not in your iCloud Backup, unless you turn that on yourself in the Settings app under Snicksnack. If you turn it on, an encrypted copy is stored by Apple, never by us.
• Wherever you choose to share or copy them.

If you delete a recording in the app, its audio and text are deleted. If you delete the app, everything is gone.

Summaries

Summaries are made by Apple's language model running on the phone (Apple Intelligence, Foundation Models). The text is not sent to Apple's servers, not to Private Cloud Compute either, and not to us. Summaries are saved together with the recording.

If you have added a receiver that offers summaries or chat, you can choose to let the receiver's language model do it instead. Then the text is sent there. Read more under Receivers.

Receivers: sending recordings to a server you choose

You can add a receiver: a server that accepts recordings, for example your employer's, a coach's or one you run yourself. You do this by scanning a QR code or entering an address and a key. So the receiver is chosen by you, or by someone you trust enough to give you a key. The app always shows which address receives.

Nothing is sent automatically. You send one recording at a time, from the Send menu. The text goes with it: the transcript with words, times and speakers, summaries and your note. If the receiver accepts audio, and you have not opted out, the audio file goes with it too.

Some receivers offer summaries and chat with a language model on their side, for example Anthropic or OpenAI through the receiver's own account. If you choose such a summary, or write in the chat, the text (not the audio) is sent to the receiver, which passes it on to the language model. The first time, the app asks clearly whether that is okay. While recording, you can choose to send what has been said so far to the chat.

When you add a receiver, the app also fetches a small settings file from its address, with a name, color and icon, and now and then it reloads the receiver's list of destinations. The receiver's server then sees your IP address, like any web server.

What the receiver does with what you send is up to the receiver, not us. Read its own terms; it stores and deletes recordings as it decides. Do not send anything you do not want the receiver to have. You can remove a receiver in the app at any time, and nothing more is sent there.

I (Mattias Ahlström) maintain an open reference receiver, snicksnack-mottagare, which anyone can set up in their own Cloudflare account. There is no central server that all recordings go to. Nothing is sent to me, unless you have added a receiver that I run.

This website

This website uses no cookies, no analytics and no external services. The server may technically log requests, like all web servers.

Contact

Snicksnack is developed by Mattias Ahlström. Questions about privacy? Email me.

This is a translation. If it differs from the Swedish version, the Swedish version applies.