Privacy

Where your voice goes — the part people do not expect

We never receive your audio. The microphone is opened by your browser, and your browser turns the speech into text.

In Chrome and Edge, that conversion happens on the browser maker's servers. Your audio leaves your device and goes to them, under their privacy policy, not ours. Safari may do it on the device. We cannot tell which, and we will not pretend otherwise — so assume your audio is leaving your device unless you know your browser does it locally.

What reaches us is the text. You can see exactly what that text was: it is printed on the page under "What it heard", every time, before anything is sent.

When the microphone is open

Only while you are holding the button or the space bar. Releasing it closes the microphone. Switching to another tab closes it too.

There is no wake word and no background listening. The page cannot hear you when you are not holding the button, and it says "Not listening" when that is the case.

It does not know who you are by voice

There is no speaker recognition here. The page cannot tell you from someone else standing nearby. Do not use it to approve anything that matters.

What we keep

At the time of writing, nothing. No database is attached to this deployment, and every reply says so. When you close the tab, the exchange is gone.

When storage is built, keeping a transcript will be something you switch on, not something that is on by default. A record of what someone said out loud in their own home is not a thing to collect quietly.

What is sent to the assistant

If you are signed in, the text is forwarded to Sova at ai.sovasta.com with a signature valid for sixty seconds that names this site and your account. If you are not signed in, nothing is sent and the page tells you so.

Asking us

support@sovasta.com. Operated by PrimeAlpha Research.