QuietTake

How it works

The tool itself is on the front page — this is everything around it: what the four steps actually do, why there is nothing to pay, what it will not fix, and the two capabilities that do not exist yet.

How it works

Four steps, none of them an upload

This is the noise removal, start to finish — the one tool of the three that exists today.

  1. Pick a file

    Drop in audio or video. Your browser reads it where it already sits — there is no upload step, so there is nothing to wait for and nothing to trust us with.

  2. The engine comes to you

    About 9 MB of WebAssembly, fetched only once you actually pick a file. This page itself stays light; the noise removal model is what arrives at that point.

  3. Cleaned on your own device

    We take a representative 30 second window and run DeepFilterNet 3 over it locally. On a recent phone that is a few seconds of work — your hardware, not our servers.

  4. Compare, then download

    Switch between the original and the cleaned version on one player, try a different preset as often as you like, and download the preview.

The whole file goes through the same four steps — same device, no upload, also free

Pricing

Free. There is nothing to buy.

You find out whether this works on your recording without any money involved, because the preview costs us nothing to give you — it runs on your hardware.

Everything, today

Free

Available now

  • A 30 second cleaned preview
  • Every preset, switched as often as you like
  • As many files as you want
  • No account, no card, no email
  • Runs on your device — your file is not uploaded
Clean a file

Also free

Cleaning a complete file

The whole file is free too, and it runs the same way the preview does — on your own device, with nothing uploaded. There is no length limit we impose; what limits it is your own hardware and how long you are willing to wait.

We may charge for something later. If that ever happens it will be for something that costs us money to run, it will be obvious which mode you are in, and no price appears on this page before the thing is real.

Also on the roadmap

Two more things to do to an interview

Neither exists yet. They are listed because they are what we are building next — and because you should be able to tell today's tool from next year's without reading a changelog. Each card says whether it is built and where your file would go.

Anonymise

Not built yet

Make a speaker unrecognisable — the voice first, the face if you need it.

For user research · documentary · whistleblower interviews

Meant to run in your browser How anonymising would work

Dub

Not built yet

Transcribe, translate and re-voice a recording in another language.

For multi-speaker interviews · podcasts · courses

Will need to upload your file How dubbing would work

FAQ

The questions worth answering honestly

Does my file get uploaded?

No — not by the noise removal, which is the only tool here that exists. Your browser opens the file locally and runs it on your own device, so there is no upload step to opt out of. That also means your audio is never used to train anything: we could not, because we never receive it. The dubbing tool on the roadmap would have to upload, but it is not built.

Then why does it download about 9 MB?

Because the engine has to come to you rather than your file going to it. That download is the noise removal model itself, and it only starts once you pick a file — landing on this page does not fetch it.

Which files can I use?

Audio or video: mp4, mov, m4a, mp3, wav, aac and most things alongside them. With a video we take the audio track and leave the picture alone. If your browser cannot decode a particular file, converting it to mp3 or wav first almost always works.

Why only 30 seconds?

Because it is enough to hear whether the noise is gone and whether the voice survived it, on your own recording rather than on a demo we chose. If it sounds right, the button under it cleans the complete file — also free, also on your device.

Is it free?

Yes, and there is nothing to buy: no account, no card, no email. That includes cleaning the complete file, not just the preview — it runs on your own hardware, so it costs us nothing to let you do it. We may charge for something later, but it would have to be something that actually costs us money to run, and no price appears anywhere before the thing is real.

Will it always sound better?

No, and you should not take our word for it either way — that is what the A/B player is for. Noise removal is lossy: the stronger settings in particular can leave a voice sounding thin or metallic. If the balanced preset already sounds right, use it.

Which browsers work?

We test in Chrome and Safari, including Safari on iPhone, where a 30 second preview takes a few seconds. Other Chromium browsers such as Edge use the same engine as Chrome. We have not tested Firefox yet.