Nothing kills an otherwise great take like a bed of hiss, room rumble or a distant air-con hum sitting under the words. The good news is that you can remove background noise from vocals with a repeatable chain of tools, and most of it can be done cleanly without making the voice sound thin or underwater. This guide walks through the whole process in order, from prevention at the source to surgical spectral repair, with real starting settings you can dial in today.
Part of our Mixing & Mastering guide — explore the full guide →
Short on time? Run a quick pass through our free online noise reducer for steady hiss and hum, or try the AI audio enhancer on spoken or sung takes, then use the steps below for anything those don’t catch.
Start at the source, not the plugin
The cleanest fix happens before you ever open a DAW. Every decibel of noise you avoid recording is a decibel you never have to fight later, and aggressive noise removal always costs you a little vocal detail. Before the next take, run this quick checklist:
- Kill the obvious sources. Switch off fans, air conditioning, fridges and buzzing chargers. A laptop fan two feet from the mic is a common culprit.
- Get closer. Moving from 12 inches to 5–6 inches lifts the vocal well above the room noise floor, improving your signal-to-noise ratio by 6dB or more.
- Tame the room. Duvets, a thick curtain or a reflection filter behind the mic reduce reflections and background bleed.
- Set gain sensibly. Aim for peaks around -12dBFS to -6dBFS. Under-recording and then boosting later drags the noise floor up with the voice.
If you are still setting up a space, our walkthrough on home recording tools and calculators can help you plan mic distance and gain staging before you commit.
The four-stage cleanup chain
Once the take is recorded, process in this order. Doing it out of order (for example, EQ-ing before you remove broadband hiss) makes each stage work harder and sound worse.
1. High-pass filter first. Almost no useful vocal content lives below 80Hz, but plenty of rumble, footsteps and HVAC hum does. Insert an EQ such as FabFilter Pro-Q and set a high-pass at 80Hz with a 24dB/octave slope for most singers, or 100Hz for a higher voice. This alone clears a surprising amount of low-end junk.
2. Notch out steady tones. Electrical hum sits at 50Hz or 60Hz (depending on your country’s mains supply) plus its harmonics at 100/120Hz, 150/180Hz and so on. Add narrow, deep notches (Q around 8–10, cut 12–18dB) at those exact frequencies. A dynamic EQ or a dedicated de-hum module is gentler than static notches if the hum comes and goes.
3. A noise gate or expander for the gaps. A gate silences the spaces between phrases where noise is most audible. Start with a threshold just above the noise floor, an attack of 1–2ms so you don’t clip word onsets, a hold of 20–40ms, and a release of 150–250ms so tails breathe naturally. Set the range/floor to around -12dB rather than full mute, so the gate ducks the noise instead of chopping it — full closes sound unnatural. An expander does the same job more transparently.
4. Broadband and spectral repair last. Gates only help in the silences; broadband hiss under the actual singing needs a spectral tool. This is where dedicated restoration software earns its keep.
The best plugins to remove background noise from vocals
Three tools cover almost every real-world case, from a light denoise to rescuing a badly recorded take.
- iZotope RX is the industry-standard restoration suite. Its Spectral De-noise module learns a noise profile from a quiet section (highlight a gap, click “Learn”), then subtracts it. The Voice De-noise, De-hum and Mouth De-click modules are tailored for speech and singing. It also has a spectral editor where you can paint out one-off noises — a chair creak or a phone buzz — like a healing brush in Photoshop. It is the most powerful option and also the most expensive; the entry-level Elements tier covers most home users.
- Waves Clarity Vx uses a neural-network model to separate voice from noise in real time. It runs as a normal insert (no offline processing), which makes it fast for streaming, podcasts and live-adjacent work. A single “Vibe” knob does most of the work; the Pro version adds frequency bands for finer control. It can occasionally introduce a slight processed sheen when pushed hard, so use the least amount that gets the job done.
- A humble gate plus EQ is genuinely enough for mild cases, and it is free in every DAW. Do not reach for a heavy denoiser if a high-pass and a gentle expander already get you there.
Whichever you choose, the golden rule is restraint. Denoisers work by removing signal, and every model guesses at what is “noise” versus “voice.” Over-processed vocals lose air, sound lispy, and develop a watery, warbling artifact in the reverb tails. Pull the reduction back until you can just start to hear the noise creep in, then ease off slightly.
Practical workflow tips
- Denoise before compression. Compressors raise quiet detail, which includes the noise floor. Clean first, then compress, or you will amplify the very hiss you are trying to hide.
- Work in a solo loop with the reverb off. Effects and a full mix can mask noise you will still hear on headphones. Judge the raw cleaned vocal, then check it in context.
- Use a noise profile from real silence. Learn-based denoisers need a genuine noise-only section — ideally a couple of seconds of room tone recorded at the top of the session. Capture that habitually.
- Automate rather than over-process. If only the verses are noisy, clean those sections harder and leave the loud choruses alone. A little manual automation beats a blanket setting.
- Mind the mix context. A tiny bit of residual hiss that is inaudible under a busy arrangement is not worth destroying the vocal to chase. Solve the problem the listener will actually notice.
Put together, the chain is simple: prevent what you can, high-pass, notch steady hums, gate the gaps, then denoise the rest with a light hand. Master that order and you will consistently remove background noise from vocals while keeping the warmth, air and detail that make a performance feel alive. For more on shaping the voice once it is clean, browse our other vocal production guides and the free calculators over on the tools hub.
Frequently asked questions
Can background noise be removed from vocals for free?
Free noise cleanup is possible with the high-pass filter, EQ notches and noise gate or expander built into every DAW, which handle rumble, hum and noise between phrases. Broadband hiss under the singing is harder and usually needs a dedicated spectral or AI denoiser, free or paid.
Why do vocals sound underwater after noise reduction?
A watery, underwater sound means the denoiser is removing too much and guessing wrongly about what is voice. Reduce the reduction amount, re-learn the noise profile from genuine room tone, and accept a little residual noise rather than stripping the air out of the vocal.
How do I get rid of computer fan noise in vocal recordings?
Computer fan noise is best handled by moving the computer away from the mic or behind an absorbing barrier, then singing closer to the mic. Any remaining steady fan hiss can be reduced with a learn-based spectral denoiser using a few seconds of room tone recorded at the session start.
Does a dynamic microphone pick up less room noise than a condenser?
A dynamic microphone is generally less sensitive than a condenser, so when used close it captures less room noise and reflection. That makes dynamics a common choice for untreated home rooms and podcasts, though mic distance and the room itself still matter more.
Should noise removal be done before or after editing vocal takes?
Basic editing and comping can happen first, but noise reduction should be applied before compression, EQ shaping and effects. Keeping some room tone from the session lets learn-based denoisers build an accurate profile, so avoid trimming every gap to silence before denoising.



