Privacy

The app

OpenVoice has no telemetry, no analytics and no crash reporting. Not switched off by default — absent from the codebase, with a CI job that fails the build if any crate that touches your microphone, transcripts, keyboard or history gains a path to a network client.

It makes exactly two kinds of outbound request, both named in the README: a model download you asked for, and an update check you can switch off in Settings. Switching it off means no request is made. The update check carries no identifier and no usage data.

Your transcripts stay on your machine, in a plain SQLite file at %APPDATA%\OpenVoice\history.db. Audio lives in RAM; the only copy that touches disk is a temporary WAV handed to the speech engine and deleted the moment the decode returns. Turning on Keep recordings, which is off by default, writes them to %APPDATA%\OpenVoice\audio instead, where they are cleared after a week.

This website

It stores nothing at all. No database is configured for this deployment, so page views are counted into a void and discarded: there is no subscriber list, no message log, and no analytics table. There is no cookie, no local storage, no fingerprint, and no third-party script — the request monitor on the home page counts third-party requests live in your browser, and you can watch it stay at zero.

If that changes, this page changes with it: the paragraphs below are rendered from the same setting that switches the storage on, so they cannot describe something the site is not doing.

Not the same thing

A page count on a marketing site is not telemetry in a dictation app, and this page exists partly to keep that distinction explicit. Nothing on this website observes anything you do in OpenVoice, and nothing in OpenVoice reports to this website.

Found something that contradicts this? SECURITY.md is the fastest route to someone who cares.