Dub a recording

Not built yet Will need to upload your file

This does not exist yet, and unlike everything else here it could not run on your own device. Both of those are worth knowing before you read further. The tool that does work today is noise removal, which runs in your browser and uploads nothing.

What it would do

Take a recording in one language and give you back the same recording in another, spoken rather than subtitled. Three models in a row:

• Transcribe the speech, with the speakers kept apart.
• Translate the transcript, keeping each speaker's turns separate.
• Re-voice it, timed so the result still lines up with the picture.

For multi-speaker interviews, podcasts, courses — material where a subtitle track is not enough.

This one would upload your file

It is the only thing on this site that would. We are putting that here, near the top, rather than in small print at the bottom, because the rest of the site says the opposite and says it truthfully: the noise removal has no upload step, and the pages describing it say so plainly. Those statements are about that tool. This one is different, and if you only read the homepage you would not guess it.

The reason is size, not preference. The noise removal ships to you as about 9 MB of WebAssembly and runs on your device. Speech recognition, translation and speech synthesis are three separate models, each of them substantially larger than that, and none is something we could rebuild small enough to send to a browser. Running them means running them on a server, and running them on a server means your recording goes there.

So if that is a problem for your material — and for interviews under an NDA, or a source who agreed to talk to you and not to us, it should be — the honest answer is that this tool would not be for that material, and anonymising is the one aimed at it instead.

What is missing

Three models in a row — transcription, translation, speech — none of which fits in a browser. It needs a backend we do not have yet.

That is a real gap, not a scheduling one. Everything on this site so far was built on the property that we never hold your files, which is also why there is no account system, no storage and no bill to pay for either. Dubbing needs all three. Doing it properly means deciding how long a file is kept, who can reach it, what happens when a job fails halfway, and what the thing costs per minute — before writing any of it.

No date, and no email box on this page to collect one from you: our support address only receives mail, so a form here would take your address and do nothing with it.

If you need this now

Several commercial services do dubbing today, and they all upload your file too — that part is not a shortcut we would be taking and they would not. If your recording is sensitive, the question to ask any of them is how long they keep it and whether it trains anything. If it is not sensitive, they will do a better job today than a tool that does not exist.