Your voice never leaves your computer
Echophrase runs entirely on your device. Your microphone audio is processed by AI models running locally on your GPU - nothing is uploaded, streamed, or sent to any server. Ever.We never see your voice or text
Audio recordings are deleted immediately after transcription. Transcription text stays in your clipboard or file - we have zero access to it. There is no “cloud processing” option. This is by design.
What we do NOT collect
| Data | Collected? |
|---|---|
| Voice recordings | Never |
| Transcription text | Never |
| Audio files | Never |
| AI model inputs or outputs | Never |
| IP addresses | Never stored |
| Device fingerprints | Never |
| Cookies | None |
What we do collect
We collect anonymous, non-identifiable telemetrics to make better product decisions - like which GPU backends people use, which AI models are most popular, and what the best default settings should be. None of this is tied to a user ID or account.Desktop app (Aptabase)
GPU backend, selected model, recording mode, feature toggles (paste-at-cursor, filler word removal). Sent to our self-hosted Aptabase instance - data never leaves our infrastructure.
Website (Umami)
Page views, referrer, browser type, country. Sent to our self-hosted Umami instance - cookieless, no tracking across sites.
Opt out anytime
Desktop telemetrics can be disabled in Settings > Data > “Help improve Echophrase”. One final event is sent to confirm opt-out, then all tracking stops. Disabling telemetrics does not affect app functionality in any way.Account data
If you sign in (optional - the app works fully without an account), we store:- Email (via Supabase OAuth) for account identification
- Subscription tier for feature access
- Payment info handled entirely by Paddle (we never see your card)