Remove Background Noise From Audio

How does this tool remove background noise from audio?

This noise remover uses spectral noise reduction: it analyzes your audio in short overlapping windows (2048 samples at a time), builds a profile of what the noise looks like across the frequency spectrum, and then subtracts that pattern from the whole recording. It is the same broad family of technique used in podcast and voice-cleanup workflows for years, running here entirely inside your browser rather than on a server.

The mode you pick — Voice, Music, or General — changes how aggressively the tool balances noise reduction against preserving the source. Voice mode is tuned for spoken recordings where clarity matters more than preserving every subtle texture. Music mode is more conservative about touching the signal, since aggressive reduction is far more audible and unwelcome on an instrumental or a mixed track. General mode sits between the two for material that does not clearly fit either category.

Auto vs. learned noise profile: which should you use?

Every noise reduction pass needs a profile of what the noise actually sounds like, and you have two ways to build one. Auto learns it automatically from the quietest parts of your recording, which works well when the noise is present throughout and reasonably consistent — a steady room hum or a fan running in the background the whole time.

Learn from a selection lets you mark the exact start and end of a noise-only section yourself, typically a few seconds of silence at the start or end of a take where you can hear only the background noise and nothing else. This usually produces a more accurate profile than Auto, especially when the recording has quiet passages that still contain some of the wanted signal, since Auto can end up building its profile from audio that was not pure noise.

Reduction and smoothing: the tradeoff to understand

Pushing the reduction amount higher removes more noise, but it is not free — very aggressive settings tend to introduce a thin, hollow, or "watery" quality to the remaining audio, an artifact of the spectral subtraction process itself. The smoothing control exists specifically to soften that effect, trading a small amount of noise-reduction strength for a more natural-sounding result.

In practice, the cleanest results usually come from a moderate reduction amount paired with enough smoothing to avoid the watery texture, rather than pushing reduction to its maximum and accepting whatever artifacts come with it. Use the Listen to removed noise option to hear exactly what a given setting is pulling out — if it includes clearly audible speech or musical content, back the reduction down.

What this can fix, and what it cannot

This tool is genuinely effective on steady, continuous background noise: hiss from a cheap microphone or preamp, electrical hum, a fan or air-conditioner running in the background, and general room tone. That category of noise is fairly consistent across time and frequency, which is exactly what spectral noise reduction is built to model and remove.

It is not built to separate one voice from another, remove reverb or room echo, or strip out background music playing under a conversation. Competing voices overlap the target voice across the same frequencies the tool is trying to preserve, so reducing them also degrades the voice you want to keep; reverb is not a noise profile at all but a smeared copy of the original signal. If your problem is a second speaker or a roomy, echoey recording rather than a steady hiss or hum, this tool will only get you partway there.

Common uses: podcasts, voice memos, and video audio

A few workflows come up constantly:

• Podcast episodes — clean up hiss or hum from a home setup, then use the volume booster to bring the episode to a consistent podcast loudness before publishing.
• Voice memos — remove steady background noise (traffic, HVAC, fan) from a quick phone recording before sharing or transcribing it.
• Video audio — clean up the audio track from a video before pairing it back with picture; use video to audio first if you need to extract the track from your video file.
• Field and interview recordings — reduce consistent environmental noise picked up on location before editing.

Noise removal vs. vocal removal: not the same job

It is easy to conflate these two, but they solve different problems. This tool reduces steady background noise, like hiss or hum, while leaving the main content (voice or music) intact. The vocal remover does the opposite in a sense: it separates a clear, intentional vocal performance from the instrumental it is mixed with, using AI source separation rather than spectral noise reduction. If you are trying to strip an unwanted hiss or hum from a recording, use this tool; if you are trying to isolate or remove a singing or speaking voice from a song, use the vocal remover instead.

Guides for this tool