Remove Background Noise From Audio with AI

Background noise — air conditioning, computer fans, street hum, room tone — sits underneath a recording and makes it sound amateur. This tool runs a speech-enhancement neural network over your file to separate voice from noise. Both models are hybrids rather than pure black boxes: RNNoise pairs a small recurrent network with a classical band-based noise suppressor, and DeepFilterNet3 applies deep filtering on top of an ERB filterbank. That is why they stay fast enough to run on your own machine, where a heavier cloud-only model could not. The model downloads once and is cached, so it works offline afterwards — and because inference happens locally, recordings you are not allowed to upload to a cloud service can still be cleaned up here.

How to Remove Background Noise From Audio with AI

  1. 1

    Pick an AI model

    RNNoise is small and fast and handles steady hiss and hum well. DeepFilterNet3 is a larger, more capable deep-learning model that is much better at separating speech from complex background noise.

  2. 2

    Download the model

    The AI model is fetched once and cached in your browser. Subsequent runs reuse it and work with no connection at all.

  3. 3

    Add your files

    Drop in one file or a whole batch. Processing runs locally, one file at a time.

  4. 4

    Denoise and download

    Press Denoise, then download each cleaned file as a WAV.

How it works

Both engines are AI and DSP working together. RNNoise runs a small recurrent neural network that estimates how much of each frequency band is speech, then drives a classical band-gain noise suppressor with that estimate — the same engine used for real-time voice in Discord and Jitsi. DeepFilterNet3 goes further: it splits the signal into an ERB filterbank, then applies learned complex filters to the low bands where speech lives, recovering detail a gain-only approach would smear. Both are compiled to WebAssembly and run on your own CPU, which is what keeps a neural model practical inside a browser tab.

Supported formats

WAV, MP3, M4A, AAC, OGG and FLAC. Files are decoded by your browser, so anything it can play will work here. Output is exported as WAV.

Frequently asked questions

Is my audio uploaded anywhere?
No. Everything runs inside your browser using the Web Audio API, so the file never leaves your device. You can prove it: load the page, turn off your Wi-Fi, and the tool still works.
Which model should I use?
Try RNNoise first — it is a fraction of the download and often enough for constant background noise. Switch to DeepFilterNet3 when the noise is louder, more varied, or competing with the voice.
Does it work on music?
Both models are trained for speech. On music they will suppress parts of the signal you want to keep, so they are best used on voice recordings.
Does this use AI?
Yes. Both options are neural networks trained to separate speech from noise, and both combine that network with classical DSP — RNNoise drives a band-based noise suppressor from a small recurrent net, and DeepFilterNet3 applies learned filtering over an ERB filterbank. The AI does the part that rules cannot express; the DSP keeps it fast enough to run in a browser tab.
Do I need an account?
No. The free tools work straight away with no sign-up. Creating a free account raises your daily processing limit from 20 to 60 minutes.
Is it free?
Yes. Every tool on this page is free to use, with a daily limit on total audio processed. There is no watermark and no trial period.

Other free audio tools