Remove background noise from audio
A clean voice from a noisy recording, made on your own device.
- Nothing is uploaded
- Works offline after first use
- Open source
- No watermark · limit is your device’s memory; phones hold less
Clean hiss, hum and fan noise out of a voice memo, podcast, interview or lecture recording in your browser. It is free, needs no account and adds no watermark, and your files are not uploaded: everything runs in your browser on your device. The size limit is your device’s memory, and phones hold less; give the page a short while after your first run to save itself for offline use, and then it works offline.
Done
Open the network tab. Your files never appear in it.
Add files and run the tool to see every request this tab makes.
Your files are processed on this device and never sent to a server. Reload the page to clear everything from memory.
A voice memo recorded next to a laptop fan, an interview in a café, a lecture with the projector humming, a podcast guest on a cheap headset: the voice is there, but so is a bed of noise that makes it tiring to listen to. This page runs RNNoise, a small neural network trained on many hours of recorded speech and noise, over every 10 milliseconds of the recording and keeps only what sounds like a voice.
Steady noise goes almost completely: hiss from a gain knob turned up too far, the 50 or 60 Hz hum of mains power, computer and air-conditioning fans, road and rain noise. Sudden sounds such as a door, a cough or a dog bark are reduced but not always removed, because for a moment they look like speech to the network.
Start with Strong, listen to the first minute, and switch to Medium if the voice sounds thin or watery. Medium and Light mix a share of the original back in, which keeps some of the room and sounds more natural on recordings where the noise was quiet to begin with. The file comes back in the same format at about the same bitrate unless you pick another under Save as.
How it works
- Drop one or more recordings.
- Keep Strong, or pick Medium or Light for a more natural sound.
- Press Remove noise and save the cleaned files.
In your browser, not on a server
Your browser decodes the sound with its own audio decoders. RNNoise (BSD-3-Clause, from Xiph.Org and Mozilla), a small recurrent neural network compiled to WebAssembly, then reads it 10 ms at a time at 48 kHz, estimates how much of each frequency band is voice, and turns the rest down. The file is written again in your tab: MP3 by LAME in WebAssembly, WAV and FLAC by code in this site, M4A and OGG by the browser’s own encoders. For a video, Mediabunny (MPL-2.0) copies the picture packets unchanged and adds the cleaned sound.
Compared with upload-based tools: Online noise removers upload your recording to their servers, cap free use at a few minutes, and often ask for an account before you can download. Interviews, voice notes and meetings are private. Here the network runs on your device and the sound never leaves it.
Verify it yourself. Give the page a short while after your first run to save itself for offline use, then turn off Wi-Fi or switch to airplane mode and run the tool again. Or open your browser's developer tools (F12, or Cmd-Option-I on a Mac), pick the Network tab, and watch it while you convert. The panel above does the same count for you.
Questions
Will it clean up a podcast with two people?
Yes. It keeps every voice it hears, whoever is speaking, and removes the background around them. It does not separate or balance the speakers; if one is much quieter, run the result through the volume booster with Normalize afterwards.
Can it remove echo or room reverb?
Only a little. RNNoise is built for background noise, not for the reflections of a bare room, so an echoey recording stays echoey. Recording closer to the microphone, or in a room with soft furnishings, is the fix for reverb.
Should I save as MP3 or WAV?
Keep the original format for sharing or listening. If you will edit the recording further, for example cut it in an editor or add music, choose WAV or FLAC under Save as, so the next export does not compress the sound a second time.
How long does it take?
Roughly a minute of work for every ten to fifteen minutes of stereo sound on a recent laptop; a mono voice memo takes half as long, and a phone takes longer. Nothing is uploaded, so the time does not depend on your connection, and the progress bar shows how far it has got.
Is the file uploaded anywhere?
No. There is no server in this process. The page is a static file and the work is done by code running inside your browser tab. Give the page a short while after your first run to save itself for offline use; then you can turn off Wi-Fi and it still works.
More video and audio tools
Remove background noise
Keep the voice, lose the hiss, hum and fan noise. Cleaned on this device, never uploaded.
Pitch and speed changer
Change the key of a song, or play it faster or slower, keeping the other the same. Done on this device.
BPM and key finder
Drop an audio file and read its tempo, key and Camelot code. The song is analysed on this device.
Online tuner
Play a note and see how sharp or flat it is. Guitar, bass, ukulele, violin or chromatic, with the sound kept in your tab.