Remove background noise from audio with AI
Background noise ruins good recordings: the hum of a fridge in a podcast, traffic in a voice memo, the fan of a laptop in an online lesson, wind in a phone video. Classic noise gates and filters either leave the noise in or make the voice sound underwater. AI noise reduction works differently. A neural network trained on thousands of hours of speech and noise listens to every 10 milliseconds of the recording, decides which parts of the sound belong to the voice and which to the noise, and keeps only the voice.
This tool uses DeepFilterNet3, an open speech enhancement model by Hendrik Schröter and colleagues, released under the MIT and Apache 2.0 licences. It is small enough to run in a web browser on any modern computer or phone, and fast: a ten-minute recording is usually cleaned in under a minute.
How to remove background noise
- Drop an audio file (MP3, WAV, M4A, FLAC, OGG) or a video (MP4, MOV, WebM) into the box above.
- Wait a moment while the noise is removed. The first run downloads the 17 MB model once.
- Press play and switch between Original and Cleaned to hear the difference.
- Adjust the noise reduction strength if you want to keep a little ambience, then download the clean audio, or the video with the clean audio.
Noise this tool removes well
Hum, hiss and electrical noise
Mains hum, buzzing cables, microphone hiss and the noise floor of cheap recorders are removed almost completely, because they are steady and very different from speech.
Fans, air conditioning and computers
Broadband noise from fans, heaters, projectors and laptops is the most common problem in home recordings and online meetings. The model removes it while keeping the voice full.
Traffic, wind and outdoor noise
Street noise, passing cars, rain and wind rumble are reduced strongly. Very loud gusts that cover the voice cannot be fully undone, but they become much less distracting.
Room noise, clicks and murmur
Keyboard typing, clattering dishes, distant chatter and general room tone are reduced. Other people speaking close to the microphone are voices too, so they are kept.
Who needs to remove background noise?
- Podcasters recording at home or remotely, where every guest has a different room.
- Video creators filming outdoors or with a camera microphone.
- Teachers and students cleaning lecture recordings and online lessons.
- Journalists and researchers working with interviews recorded in cafés, streets and offices.
- Anyone with a voice memo or a voicemail that is hard to understand.
Tips for cleaner results
- Do not over-clean. For videos, 70 to 90 % noise reduction often sounds more natural than 100 %, because a little room sound hides small artifacts.
- Level the result. Tick "Even out loudness" to bring quiet recordings up to a standard podcast and video level, or use the audio enhancer, which turns it on for you.
- Music is not noise. The model is trained to remove noise, not songs. To take music out from behind a voice, use the background music remover or the voice isolator.
- Start from the original file. Messaging apps compress audio heavily; the original recording cleans up better.
Noise remover, voice isolator or music remover?
| Problem | Best tool |
|---|---|
| Hum, hiss, fans, traffic behind speech | Remove background noise (this page) |
| Quiet, uneven or noisy voice recording | Audio enhancer |
| Music playing behind speech | Background music remover |
| Voice buried in music and noise | Voice isolator |
Private and free
Nothing you drop here is uploaded. The noise reduction model is downloaded to your browser once and cached, and all processing happens on your own device in WebAssembly. That also means it works on Chrome, Edge, Firefox and Safari without a graphics card. There is no sign-up, no limit on files and no watermark.