How it works

The speech model runs on your computer, inside the browser tab. Here is what happens to your file, and how to check it yourself.

What happens when you drop in a file

1. Your browser opens the audio.
2. It finds the parts where someone is talking.
3. Only those parts go to the speech model. Silence is skipped.
4. Text appears as it goes, with timestamps.
5. A final check flags any text stuck repeating itself.
6. The text and the recording are kept in this browser, on this computer.

Why silence is skipped

Speech models sometimes invent a sentence when they hear silence or noise. Every model does this. So Ratproof skips silence and flags text that loops. It does not catch every mistake. See the numbers.

Check it yourself

Watch the network. Open Ratproof, press F12, and open the Network tab. You will see the page load, plus the speech model on your first visit. Then drop in a recording. Entries that start with blob: are the page reading your file from memory. No new entry goes to another server.

Your browser enforces it. The page tells your browser to block the page's own code from connecting to any other server. You can see that rule in the response headers.

No tracking. No analytics, no usage data, no outside scripts. Even the font comes from its own server.

Everything it is built on is open source.

Want it in its own window, working without internet? Install it from Edge or Chrome.