Privacy

The short version

Mimic collects nothing about you. There is no sign-up, no profile, no advertising, no analytics and no tracking of any kind. Speech is generated by a model running on your phone, which is why the app is large and why it works in aeroplane mode.

Your voice and your recordings

When you record yourself reading a paragraph, that recording is written to Mimic’s own storage on your phone and used to build a voice profile there. Neither the recording nor the profile is ever uploaded. If you export a voice to move it to another device, the file contains the recording, and where it goes after that is up to you. Deleting a voice in the app deletes its recording with it, and deleting the app removes everything at once.

What you type, and what Mimic says back

Passages you write are kept on the phone so a draft survives closing the app. Audio Mimic generates is rendered on the phone and cached there, so that saying the same thing twice does not cost the wait again. Both are yours to clear: the cache has a button in Settings, and a draft is gone when you replace it.

The one thing that leaves your phone

The speech model is not bundled with the app — it would be far too large — so on first run Mimic downloads roughly 600 MB of model files from Hugging Face, another 400 MB the first time you record a voice, and, only if you ask for the optional writing model, about 468 MB more. These are ordinary HTTPS requests for public files. As with loading any web page, Hugging Face and its content delivery network can see your device’s IP address and which file was requested. Nothing about you, your voice, your recordings or your text is sent, because none of it is involved in asking for a file.

That is the entire list. Mimic makes no other network requests. Links in Settings to the model authors and to the source code open in your browser, which is then governed by your browser and those sites.

Writing a passage for you

Write me one asks a language model to draft something to say. If your phone has Apple Intelligence available, Mimic uses Apple’s model, which runs on the device; Apple’s own privacy policy governs it. Otherwise Mimic can download a small model and run it itself, on the phone. In both cases your prompt stays on the phone.

Permissions Mimic asks for

• Microphone — only to record you reading a paragraph, so a voice can be made from it. Mimic does not listen at any other time.
• Adding to your photo library — only when you choose to save a passage as a video. Mimic is never given permission to read your library.

Both are optional and asked for at the moment they are needed. Refusing either leaves the rest of the app working.

Children

Mimic is not directed at children. It collects no personal information from anybody, whatever their age.

Changes

If a future version of Mimic does something this page does not describe, this page changes first, and the date above changes with it.

Asking about any of this

Write to david.cvetkovski@gmail.com. The app is open source, so the claims on this page can be checked rather than taken on trust: github.com/DavidCvetkovski/mimic.