Most privacy pages are long because they have to be. Ours is short because there's no backend.
Audio buffer, live waveform, transcribed text, history, dictionary, saved snippets, preferences.
There is no WisperType server that audio or transcripts could go to. There is no analytics SDK embedded.
While you hold the hotkey, your microphone feeds a rolling buffer in RAM. That buffer is passed to Whisper, which runs locally via whisper.cpp. As soon as the transcription returns, the audio is discarded. It is never written to disk and never sent over a network.
Transcribed text is written to a local SQLite database inside
~/Library/Application Support/WisperType.
You can see everything in the Home view, delete individual entries,
or enable auto-delete after 24 hours. There's no sync service.
None. No analytics SDK, no product-usage tracking, no crash reporting without your explicit opt-in. If the app crashes, you decide whether to send us the stack trace.
Subscriptions are handled by Paddle. They process your email and payment method, not your audio, not your transcripts. That's the only data that ever leaves your Mac, and only because payment card networks require it.
The Whisper model is downloaded once from our CDN when you pick a model size. App updates are fetched from our release endpoint on launch. Both are vanilla HTTPS file downloads — no tokens, no tracking pixels.
If your Mac is air-gapped, WisperType still works — that is the clearest statement of intent we can give you.
Last updated: April 2026. Questions? privacy@wispertype.com.
Install WisperType and never wonder where your voice went again.