Free tool
Remove background noise. With a neural denoiser, free, in your browser.
How it works
01
Drop in a recording
A voice memo, podcast, video voice-over or vocal take: WAV, MP3, M4A or FLAC, up to ten minutes.
02
The network cleans it
DeepFilterNet 3, an open neural speech-enhancement model, runs on your device and separates the voice from hiss, hum, fans, traffic and room noise.
03
Compare and download
Listen before and after, then save a WAV. Choose Gentle if Full sounds too dry.
Questions
- What noise does it remove?
- Steady and changing background noise around speech or singing: hiss, hum, air conditioning, computer fans, traffic, keyboard clatter and room tone.
- Does it work on music?
- It is trained on speech, so it treats instruments as noise. Use it on voice recordings and vocal takes, not finished songs.
- What's the difference between Full and Gentle?
- Full removes as much noise as it can. Gentle caps the reduction at 18 dB, which keeps a little of the room and sounds more natural on some voices.
- Is my audio uploaded?
- No. The model is downloaded once to your browser and the processing happens on your device. The first run takes a moment longer while it loads.
- Is it free?
- Yes, with no sign-up and no limit beyond ten minutes per file.
Clean takes.
Now produce them.
The same denoiser runs on any clip in the SoundGen studio, next to a full mixer, effects and an AI producer.