When audio is captured
We only listen while you’re actively using a voice-entry control (for example a mic button on a field). Ending the session or leaving the screen stops capture. We don’t keep the mic “warm” in the background for advertising or analytics.
On-device vs cloud processing
The operating system or bundled speech engine may process audio entirely on-device to produce text. If a future feature requires cloud transcription, we’ll label that clearly before you opt in, describe what leaves the phone, and name the provider. Today’s default posture: treat voice as sensitive workshop chatter until a product changelog says otherwise.
Retention
Raw audio isn’t a souvenir. Unless a bug splinters that rule, we don’t store long-lived sound files server-side from dictation — we store the text you confirm into your records, which then follow normal sync and retention rules described in Data practices.
What we don’t do
- No always-on ambient recording.
- No selling voiceprints or biometric voice models.
- No training generalized AI models on your customers’ conversations without a separate, explicit agreement.
Your controls
Deny microphone permission in system settings and the app falls back to typing. You can review transcribed text before saving, same as any other field — muttering isn’t destiny.
Related: Privacy Policy · Data practices · Camera & photos