Does it work on music?

No. It is built for speech, and it treats music, singing and other non-speech sound as noise to remove. To separate a voice from music, the right tool is a vocal remover, not a noise remover.

Can it remove echo, or a second person talking?

No. It is not built to remove echo, so a recording made in a bare, echoing room stays roomy. Another voice talking at the same time is speech too, so it is kept.

How is this different from Adobe Podcast Enhance?

Adobe's tool uploads the recording, needs an account, and limits free use to an hour a day and 30 minutes a file. It also rebuilds the voice, which can change how it sounds. This one removes the noise and keeps the voice that was recorded, on your own computer, with no account and no limit.

How long does it take?

On a typical laptop the best-quality cleaner runs faster than the recording plays, and long files are split across several processor cores. On a recent computer an hour-long podcast takes a few minutes. Firefox runs it more slowly than Chrome, Edge or Safari. Fast uses a smaller network and is quicker still.

Which files can it open?

Most audio (MP3, M4A, AAC, WAV, FLAC, OGG, Opus) and video (MP4, MOV, WebM, MKV) that your browser can play. A video keeps its format; a WebM whose sound your browser cannot encode is saved as MKV. For other formats, convert first with the Video Converter.

Does the recording leave my computer?

No. It is decoded, cleaned and saved by your browser, on your computer. Only the noise model is downloaded, once, from this site.

Stays on your computer How this works

Remove background noise from audio and video

Clean up voice recordings, podcasts, interviews and videos: hiss, hum, fans, traffic and room noise taken out with an AI model that runs on your computer. Free, with no upload and no length limit.

Drag & drop audio or video here, or browse

MP3, M4A, WAV, FLAC, OGG, MP4, MOV, WebM and more · several at once · nothing is uploaded

Cleaner

DeepFilterNet 3: a 10 MB download the first time.

Strength 100%Lower keeps a little of the room, which can sound more natural.

Enhance voiceCut low rumble and set the level to -16 LUFS, the usual loudness for podcasts.

Save audio asVideos stay videos: only their sound is replaced.

About removing background noise

This tool takes the noise out from under speech: the hiss of a cheap microphone, mains hum, a fan or air conditioner, traffic outside, keyboard clatter and the general murmur of a room. It uses DeepFilterNet 3, an open-source speech-enhancement network that works in the same way as the paid AI cleaners: it hears which parts of the sound are voice and keeps them, and turns everything else down. The network runs on your computer, in the page, so recordings are never uploaded and there is no limit on how long they are or how many you clean. It is free to use, with no account and no upload limit, and it runs entirely in your browser: nothing is uploaded, and while the page stays open, a feature you have already used keeps working offline.

It works on video too. The sound is taken out of the video, cleaned, and put back, while the picture is copied across untouched, so the video loses no quality. Enhance voice then cuts low rumble and sets the loudness to -16 LUFS, the level podcasts and YouTube voiceovers are usually mixed to, without letting any peak clip. Listen to the original and the cleaned version side by side before you save.

What you can do

Remove background noise, hiss, hum and room noise from voice recordings
Clean the sound of a video without re-encoding the picture
Choose how strongly noise is removed, from a light touch to all of it
Level the voice to podcast loudness and cut low rumble with Enhance voice
Compare the original and the cleaned sound at the same moment
Clean several files at once and save them together as a .zip
Save audio as WAV, MP3, M4A or FLAC, or in the format it came in

How to remove background noise from a recording

Drop the audio or video file onto the page, or click to choose it. Several can go in at once.
Wait while it is cleaned. The first time, the noise model (about 10 MB) is downloaded; after that it is quicker to start.
Switch between Original and Cleaned while it plays. If the voice sounds too processed, lower the strength and clean it again.
Press Save. A video comes back as a video with the clean sound.

Questions

Does it work on music?: No. It is built for speech, and it treats music, singing and other non-speech sound as noise to remove. To separate a voice from music, the right tool is a vocal remover, not a noise remover.
Can it remove echo, or a second person talking?: No. It is not built to remove echo, so a recording made in a bare, echoing room stays roomy. Another voice talking at the same time is speech too, so it is kept.
How is this different from Adobe Podcast Enhance?: Adobe's tool uploads the recording, needs an account, and limits free use to an hour a day and 30 minutes a file. It also rebuilds the voice, which can change how it sounds. This one removes the noise and keeps the voice that was recorded, on your own computer, with no account and no limit.
How long does it take?: On a typical laptop the best-quality cleaner runs faster than the recording plays, and long files are split across several processor cores. On a recent computer an hour-long podcast takes a few minutes. Firefox runs it more slowly than Chrome, Edge or Safari. Fast uses a smaller network and is quicker still.
Which files can it open?: Most audio (MP3, M4A, AAC, WAV, FLAC, OGG, Opus) and video (MP4, MOV, WebM, MKV) that your browser can play. A video keeps its format; a WebM whose sound your browser cannot encode is saved as MKV. For other formats, convert first with the Video Converter.
Does the recording leave my computer?: No. It is decoded, cleaned and saved by your browser, on your computer. Only the noise model is downloaded, once, from this site.

Built by John McMillan, a senior full-stack engineer. Open to remote roles: hire me.

Report a problem or suggest a feature