Quick answer: It never leaves your device. Audio is processed entirely on-device using an ML model that ships with the app. We cannot hear your recitation — even if we wanted to.
How on-device verification works
When you recite into the microphone, the audio is fed directly into a compact ML model that runs on your phone’s CPU (or Apple’s Neural Engine / Android’s NNAPI when available). The model outputs “word correct / incorrect / which sub-rule” — those outputs are what feed into your retention score. The raw audio is never uploaded, cached to disk (except briefly during playback), or transmitted to any server.
What we can verify
- The app requests microphone permission only when you start a verifier session.
- The app does not send microphone data over the network. You can verify this by using network-monitoring tools like Little Snitch (Mac) or NetGuard (Android).
- Our privacy policy is auditable — email security@hafizprime.com to arrange a technical review.
What we do store
- Your retention scores per ayah — a single number.
- Tajweed areas flagged as weak (e.g., “qalqala”).
- Timestamps of verifier sessions.
These are aggregated and shared only with teachers or parents you’ve explicitly linked.
Deleting audio
Since we never receive your audio, there’s nothing server-side to delete. The audio buffer on your device is cleared at the end of each session. If you revoke microphone permission, even the local buffer is blocked.
