Anonymise a speaker

Not built yet Meant to run in your browser

This does not exist yet. Nothing on this page is something you can use today — it is here so you can tell what it would do and what is actually missing, rather than finding it in the navigation and wondering. The one tool that does work is noise removal, and it works now, in your browser.

What it would do

Make a speaker unrecognisable in a recording, in two halves that are really two different jobs:

• The voice. Shift pitch and formants enough that the speaker cannot be identified by ear or matched against a voice sample, while leaving the words intelligible. Not a robot filter — those are unpleasant to listen to for more than a minute, which defeats the point if anyone has to actually watch the interview.
• The face, if the recording has picture. Find the face across the shot and cover or blur it, without needing you to keyframe it by hand.

For user research, documentary, whistleblower interviews — recordings where the content matters and the identity must not.

The voice half is the tractable half

Pitch and formant shifting is signal processing, the same category of work as the noise removal that already runs here. That matters practically: it means the same route we have already proven — a WebAssembly engine that runs on your own device — and not a new class of problem. We are not claiming it is done. We are saying we know what it is.

The face half is not

Face detection in a browser is possible, and there are models that do it. We have not picked one, measured it, or found out what it costs on a phone — and until we have, any number we gave you about it would be invented. This is the half that decides whether the tool ships as voice-only first or not at all until both work.

Where your file would go

The intention is your own device, the same as the noise removal: your browser opens the file where it sits, and nothing is sent to us. That is why the badge above says meant to rather than does. It is not built, so it cannot promise anything about where it runs — a tool that does not exist cannot keep a privacy commitment, and stating one anyway is the kind of claim that is easy to make and impossible to check.

Worth being explicit about the site as a whole: noise removal runs locally and always has. Dubbing, the third thing on the roadmap, would have to upload your file — see how dubbing would work. So "nothing is uploaded" is true per tool, not for the site, and each tool says which it is.

What is missing

The voice side is a known signal-processing job; the face side needs a detector we have not picked yet, and neither is written.

No date. We are not going to give you one for something where half the work has not been scoped, and a date we invented would be worth less than this sentence. There is also no email box on this page on purpose: our support address only receives mail, so a form here would take your address and then do nothing with it.

If you need this now

Use something that exists. Voice changing is available in several audio editors, and face blurring in most video editors — both as manual steps rather than one action, which is exactly the gap this tool would close. We would rather point you at a real workflow than have you wait on ours.