Background Noise Remover

Remove background noise from audio, free and without uploading

This background noise remover takes the hum, hiss and room noise out of a voice recording and leaves the voice. Choose a file, wait while it is cleaned, then switch between the original and the cleaned version on the same timeline to hear the difference before you download anything.

It all happens in your browser. The recording is never uploaded, there is no account to make, no watermark, and no limit on how many files you clean. The model behind it is DeepFilterNet3, an open source speech model that has been trained to tell a voice apart from the noise around it.

How to use it

1. Choose an audio or video file, or drop it on the box above. Cleaning starts right away.
2. The first time, the noise model downloads (about 17 MB). After that it loads from your device.
3. Press play and switch between Original and Cleaned. Lower the Strength if the voice sounds too dry.
4. Download the WAV, or for an MP4 or MOV, the video with the cleaned sound.

Remove background noise from video too

Phone videos, screen recordings and interviews filmed in a busy room are where noise is worst. Choose the video itself: only its sound is read, the picture is never decoded, so cleaning a long phone video is fine. For MP4, MOV and M4V files up to 750 MB (300 MB on a phone) the cleaned sound is put back into a new MP4 with the picture copied over exactly as it was, ready to post or edit. For a larger video, download the cleaned WAV and put it back in your editor.

What it removes well, and what it does not

• Removes well: fan and air conditioning noise, computer and electrical hum, mic hiss, traffic, rain, keyboard noise, the murmur of a café or office.
• Removes, though you may not want it to: background music. It is treated as noise, so a music bed under a voice-over mostly disappears.
• Does not fix: echo and reverb, two people talking over each other, and speech that was distorted or clipped when it was recorded.

Why it runs on your device

Most online noise removers upload your recording, clean it on their servers, and ration how much you can do for free, because every minute costs them computing time. Here the work happens on your own device, which changes three things:

• Privacy. Your audio never leaves this device. Calls, interviews and voice notes stay yours.
• Cost. There is no server bill to pass on, so it is free, unlimited and needs no signup.
• Offline. Once the model is saved, it works without a connection.

Frequently asked questions

Is my recording uploaded anywhere?

No. The file is read and cleaned inside this browser tab, and the cleaned audio is saved straight from it. There is no server in the loop, nothing is uploaded, and there is nothing for us to log or keep.

How do I remove background noise from a video?

Choose the video (MP4, MOV, MKV or WebM). The sound is pulled out of it and cleaned on your device. For MP4, MOV and M4V files up to 750 MB (300 MB on a phone), Download video saves a new MP4 with the picture copied over untouched and the cleaned sound in place of the old one. For larger videos, MKV and WebM, download the cleaned WAV and put it back in your editor.

What kinds of noise does it remove?

Steady and semi-steady noise around a voice: air conditioning, fans, computer hum, electrical buzz, hiss from a cheap mic, traffic, rain on a window, keyboard clatter and the general murmur of a room. It is trained on speech, so it works best when the thing you want to keep is someone talking.

What can't it fix?

It is not a music separator: background music is treated as noise and mostly removed, together with any singing. It does not take echo or reverb out of a hall or bathroom recording, and it cannot pull one voice out of several people talking at once. Speech that is clipped or distorted stays distorted.

What does the Strength slider do?

At 100% you hear only the cleaned audio. Lower blends some of the original back in, which keeps a little of the room and can sound more natural on a voice recorded far from the mic. The two are lined up sample for sample, so the blend never smears. The slider works live while you listen and applies to the downloads.

Why does it download a model the first time?

The noise removal model runs on your device, so it has to be on your device. It is about 17 MB, downloaded once and kept by your browser, so the next file starts straight away and works offline. You can see and delete it with the Manage models button at the top of the page.

How long can the recording be?

Up to 60 minutes at a time on a computer and up to 10 minutes on a phone, where the browser gets far less memory. For a longer recording, trim it into parts and clean each one.

What file do I get?

A WAV file, 16-bit at 48 kHz, which every editor, phone and messaging app opens. It is saved as mono: the model works on one channel, so a stereo recording is mixed down first. For a video, you can also save an MP4 with the cleaned sound.

Does it work on iPhone and Android?

Yes. The model runs on the processor rather than the graphics chip, so it works the same in Safari, Chrome and Firefox, on a phone or a computer. On iPhone you can pick a video from your photo library or record one from the file menu.

Which model is this, and is it free to use commercially?

DeepFilterNet3, an open source speech enhancement model by Hendrik Schröter, released under the MIT and Apache 2.0 licences. The cleaned audio is yours to use however you like, commercially included, and we claim no rights over it.