Microphone and voice data
- Capture begins only after the learner takes an explicit action. RecallSpeak does not listen in the background.
- Real Conversation requires every participant's agreement before recording. A visible recording state and stop control remain available.
- Raw audio and the short learner reference are held only for the processing request, then discarded from RecallSpeak application memory.
- RecallSpeak does not store raw audio, full transcripts, or a voiceprint in its application data.
Learning data
RecallSpeak stores a pseudonymous learning profile, minimized and redacted evidence, confirmed learning needs, review progress, and scheduling data. Real Conversation findings remain candidates until the learner confirms them. Silence, accent, or speaking speed alone is not used to infer a cause or sensitive trait.
Accounts and purchases
Supabase processes the account identifiers and session data needed for sign-in. In the mobile apps, RevenueCat processes a pseudonymous app user ID and purchase or entitlement status so RecallSpeak can determine plan access. Apple or Google may also process store transactions under their own terms.
Provider retention and controls
Voice or text submitted for analysis is processed by OpenAI. Provider-side retention depends on the exact endpoint and account settings, so deleting RecallSpeak data cannot guarantee immediate deletion by every provider. The release build includes an in-app account-deletion control; store subscriptions must be cancelled separately with Apple or Google.
