Noise Remover

Remove Background Noise from Video and Audio

Removes fan hum, traffic, wind, keyboard and room noise from speech. It runs in your browser, so the file is never uploaded.

Hear it in 10 seconds Fan noise, real voice. Press play, then flip the switch.
0:00 / 0:11
  • Free preview of any 30 seconds, with an Original and Cleaned switch.
  • Private. Your video or audio never leaves your device.
  • Your controls. Strength slider, hum filter and volume levelling.

No file handy?

Hear the difference

Real voice, with real noise mixed in for each demo, cleaned by the same engine at maximum strength. Pick a noise type, press play and switch between Original and Cleaned.

0:00 / 0:11

Fan and air conditioning. Real voice, noise mixed in for this demo.

Noise it handles

Each demo above is a real recording of that noise. What changes depends on the noise, so here is what to expect.

Fan and air conditioning

Steady broadband noise is the easiest case, and the one this model handles best.

Traffic

Road rumble and passing cars under a voice drop a lot. A horn or siren right beside the speaker can leak through.

Wind

Steady wind and low rumble drop sharply. Gusts that overload the microphone distort the voice and cannot be repaired.

Keyboard typing

Typing behind speech is reduced. A click that lands exactly on a word can leave a trace.

Cafe and crowd chatter

Background voices are turned down. Loud ones close to the microphone can still be heard.

Mains hum

50 or 60 Hz buzz is removed, and the Remove hum switch adds notch filters for stubborn cases.

Who it is for

  • PodcastsTake the fan, the street and the laptop out of remote interviews.
  • YouTube videosClean the voice track of vlogs and tutorials without re-encoding the picture.
  • Zoom and webinar recordingsMake long meeting recordings easier to listen to.
  • InterviewsRescue field interviews recorded next to traffic or a busy room.
  • LecturesReduce projector fans and room noise in classroom captures.
  • Phone and GoPro videoTurn down wind and background hiss in clips shot on a phone or action camera.

How it works

  1. Choose a file. Drop a video or audio file. The audio track is decoded on your device.
  2. Preview and adjust. Hear a free 30 second preview, move the window to any part of the file, and set the strength.
  3. Clean and download. Clean the whole file in parallel on your CPU cores, then download it. Video keeps its original picture.

The audio is decoded with FFmpeg compiled to WebAssembly, cleaned by DeepFilterNet3 in parallel Web Workers, then written back. Nothing is sent to a server.

How this compares

Editor tools (CapCut, Premiere, Audacity)Upload-based AI toolsThis tool
File leaves your deviceUsually no for desktop appsYes, it is uploaded to a serverNo, it stays in your browser
Preview before you commitYes, inside the editorUsually a short previewYes, free, no account
Choose which part to previewYes, scrub the timelineOften only the start of the fileYes, move a 30 second window anywhere
Strength controlManual settings that differ by toolOften a single on/offSlider from light to maximum
Speed and queueDepends on the app and your computerUpload time plus a server queueNo upload or queue; depends on your CPU
Install neededYesNoNo

A general comparison. Features vary by product and version.

What it can and can't fix

Works well on

  • Steady noise: fans, air conditioning, hum, hiss
  • Traffic, wind and room rumble under a voice
  • Keyboard and mouse clicks behind speech
  • Podcasts, calls, interviews, screen recordings, phone video

Limits

  • It is built for speech. Music, singing and sound effects may be dulled or partly removed.
  • It cannot separate two people talking over each other, or a voice from a TV or radio behind it.
  • Heavy clipping and very low-quality recordings cannot be restored.
  • Strong can make quiet voices sound slightly thin. Lower the strength if so.

Privacy

Your file is read, cleaned and saved inside this browser tab. Nothing from it is uploaded, and there is no account. The page downloads its own code, FFmpeg and the noise-removal model, then works without sending your media anywhere. Closing the tab clears everything from memory.

FAQ

Which file formats work, and how big can the file be?

Audio: MP3, WAV, M4A, AAC, FLAC, OGG, AIFF and Opus. Video: MP4, MOV, WebM, MKV, M4V, AVI, WMV and FLV. The limit is 1 GB and 60 minutes. Everything runs in your browser's memory, so very long files need a computer with plenty of free RAM.

Is my file uploaded anywhere?

No. The file is read, cleaned and saved inside your browser tab. The page downloads its own code and the noise-removal model, and none of your audio or video is sent over the network.

Can it clean up phone, GoPro or camera video?

Yes, as long as the noise sits behind speech. Phone, action camera and DSLR clips are handled like any other video: the audio track is cleaned and the video stream is copied unchanged, so the picture is not re-encoded.

Does it work on music?

It is built for speech. Music, singing and sound effects can be dulled or partly removed, so it is a poor choice for music recordings. If you try it, pick a low strength and compare with the original.

Can it remove wind noise?

Steady wind and low rumble drop sharply. Gusts that hit the microphone hard distort the voice itself, and that damage cannot be repaired. A windshield or a lavalier microphone still beats any fix afterwards.

Can it remove echo and reverb?

It is built to remove noise, not echo. A slightly lively room may sound a little tighter, but a very echoey room will still sound echoey.

What happens to stereo audio?

The noise remover works on one channel at 48 kHz, so stereo is mixed down to mono first. Audio files come back as mono. Video files get the cleaned mono track copied to both stereo channels.

How long does it take?

It depends on your computer. The audio is split into pieces of about 30 seconds and cleaned in parallel on several CPU cores, so a faster, multi-core machine finishes sooner. After the preview, the page shows the speed it measured, and the progress bar shows percent done and time left.

What do Remove hum and Level voice do?

Remove hum detects 50 or 60 Hz mains hum and applies narrow notch filters at that frequency and its harmonics before cleaning. Level voice normalises loudness to -16 LUFS with a -1.5 dBTP true-peak ceiling. Both are off by default.

Is payment safe?

Payment is handled by Stripe on its own checkout page, so your card details never reach this site. You can pay by card, Apple Pay or Google Pay. It is a one-time price for one file, with no account and no subscription, and you can ask for a full refund within 14 days.

What if it does not work on my file?

Preview first: any 30 seconds of your file are free, and files up to 2 minutes export free. If you paid and the result is not what you hoped for, ask for a full refund within 14 days on the refund page. You only need the email from your receipt.

Which model does this use?

DeepFilterNet3, an open-source speech enhancement model (MIT and Apache 2.0 licenses), compiled to WebAssembly and run locally on your device.

Credits

The demo clips mix a real human voice with real noise recordings, so the before and after use the same voice. Only public domain and CC0 sources are used.