Support

Reporting something broken

Email is the quickest route, and an issue on GitHub works too. Your phone model and iOS version help more than anything else, because most of what goes wrong here is about how much memory a particular phone will give an app.

Why is the download so large?

Because the model is on your phone rather than on a server. That is the whole point: Mimic generates speech without a network, so nothing you record or type has to go anywhere. The cost of that promise is about 600 MB once, and 400 MB more the first time you record a voice, best fetched over Wi-Fi. Settings shows exactly which megabytes are the speech model, which are the optional writer, which are your voices and which are cached audio — and lets you get each of them back.

The download stopped, or the app says the model is missing

Open the app again and it resumes where it left off; files already fetched are not fetched twice. If it will not finish, check that the phone has a little over 1 GB free and try again on Wi-Fi.

Why is there a wait before it speaks?

Generating speech is slower than listening to it, so there is always a pause before the first sound. Mimic starts playing as early as it provably can without running out of audio part-way through, and the countdown you see is its estimate from how fast your phone managed last time. Saying the same passage in the same voice again is instant, because it is kept.

The clone does not sound like me

Around fifteen seconds is the sweet spot, and longer is not better. Record somewhere quiet, speak at a normal level, and read the paragraph as it is written — the model is told what the reference says, and a disagreement between the recording and the text is the commonest cause of a poor result. You can keep several voices and compare them.

Does it work offline?

Once the model is downloaded, yes, entirely — including in aeroplane mode. The one exception is downloading the optional writing model, which needs a connection the first time.

Deleting your data

Everything Mimic has is on your phone. Delete a voice to delete its recording, clear cached audio from Settings, or delete the app to remove all of it at once. There is no account and nothing held anywhere else, as the privacy page sets out.

What it needs

An iPhone running iOS 17 or later, a little over 1 GB of free storage, and enough memory to hold the model while it speaks — which is the part older phones struggle with. Write me one uses Apple’s own writing model where Apple Intelligence is available, and offers to download a small one of its own where it is not.