Speak

Privacy Policy — Speak

Last updated: 31 July 2026

Speak is a pronunciation trainer for iPhone. It runs entirely on your device.

Speak does not collect any data. There is no account, no server, no analytics, and no advertising. Nothing you do in the app is transmitted anywhere.

That is the whole policy. The rest of this page explains exactly what happens to each kind of information, because “we don’t collect anything” is a claim that deserves to be checkable.

Your voice

When you tap the microphone and repeat a sentence, the audio is captured into memory, scored, and discarded.

It is never uploaded. The app makes no network requests of any kind.
It is never written to disk. The recording exists as raw samples in the app’s memory for the few seconds an attempt lasts, and is released afterwards.
The scoring is done by a speech model bundled inside the app — a 339 MB acoustic model that runs locally. That is why the download is large, and it is the reason no server is needed.

You can verify this: turn on Airplane Mode. Everything works exactly the same.

iOS will ask for microphone permission the first time you record. You can revoke it at any time in Settings ▸ Speak ▸ Microphone; scoring stops working, and nothing else changes.

Voice contributions (optional, and only if you start them)

Speak includes an optional feature for native speakers who want to help improve its pronunciation scoring: you can record a set of prompt sentences.

Those recordings are saved, as audio files in Speak’s own private folder on your iPhone. Only Speak can read them.
They stay there. Nothing is sent automatically, ever.
They leave your device only if you tap the export button, which opens your own Mail app with a zip attached and the address filled in. You see the message before it goes, you can edit or delete anything in it, and you can cancel.
You can delete individual recordings, or all of them, from inside the app at any time.

If you do email them, you are donating those clips to improving Speak’s scoring. They are used for that and nothing else. No name, email address or any other identifying detail is attached by the app — the only identity involved is whatever your own email account puts on the message you chose to send.

Your progress

Levels, streaks, XP, best scores and per-sentence mastery are stored on your device — in the app’s own settings and in a single JSON file inside its private container. None of it is sent anywhere, and none of it is readable by other apps.

Deleting Speak deletes all of it. There is no copy elsewhere, and no way to restore it, because there is nowhere for it to have been restored from.

What Speak does not do

No accounts, no sign-up, no email address, no password.
No analytics or telemetry SDK of any kind.
No crash reporting service.
No advertising, no ad identifier (IDFA), no App Tracking Transparency prompt — because there is nothing to track.
No third-party SDK except ONNX Runtime, Microsoft’s open-source local inference library, which runs the speech model on your device and does no networking.
No cookies, no web views, no embedded browser.
No location, contacts, photos, calendar, health or clipboard access.

Leaving the app

A permanent banner under the practice deck opens vzw Hoop voor Hafsa, a Belgian non-profit raising funds for a twelve-year-old girl with beta-thalassemia major. From that screen, three of the four ways to help leave for Safari: her campaign and the non-profit’s page at weglow.world, and the stem-cell donor registry at stamceldonor.be.

At that point you have left Speak, and the site you land on has its own privacy policy. Speak takes no donations, handles no money, and receives nothing back — it does not know whether you donated or even whether the page opened.

Children

Speak is rated 4+ and is suitable for all ages. It collects no personal information from anyone, including children under 13 (and under 16 in the EU), because it collects no personal information at all.

Speak processes no personal data on our side, so there is nothing for us to disclose, correct, export or erase, and no data controller relationship arises from using the app. All processing happens locally on your device under your own control.

The one exception is a voice contribution you choose to email. If you have sent one and want it deleted, write to the address below with the date you sent it and it will be removed.

Changes

If a future version of Speak ever changes what it does with data, this page changes first, and the change is called out in the release notes. A version of Speak that phoned home would be a different app, and would be described as one.

Contact

Anthony De Smet — anthony@homevote.io