Last updated: July 18, 2026
Hush is a voice dictation app that runs entirely on your device. We do not collect, store, or transmit your voice, your transcripts, or your typing.
Hush turns your speech into text using on-device speech-recognition and text-cleanup models (Zipformer, Gemma-based E2B, Parakeet). All of that processing — capturing your audio, transcribing it, and cleaning it up — happens locally on your phone. Your voice recordings and the text they produce are never sent to Hush's developer, to Google, or to any other server.
RECORD_AUDIO) — used only while you are actively
dictating, to capture your speech for on-device transcription. Audio is processed in memory and
discarded once transcription completes; it is never saved to a file or transmitted anywhere.INTERNET) — used for exactly one purpose: downloading
the on-device speech and language model files (from huggingface.co) the first time you set up
Hush. This only happens after you explicitly confirm a one-time download in an in-app consent
dialog. Hush has no analytics, advertising, or crash-reporting network calls of any kind.If you teach Hush custom words or names (its personal dictionary feature), those terms are stored locally on your device only. They are never uploaded or shared.
Hush is not directed at children and does not knowingly collect any information from children.
If this policy changes, the "Last updated" date above will change and the new version will be posted at this same URL.
Questions about this policy or Hush's data practices: dondeti.apps@gmail.com