Noisero
All guides

AUDACITY NOISE REDUCTION TROUBLESHOOTING

Audacity Noise Reduction Not Working?

Audacity Noise Reduction can work well on steady background noise, but it can also leave noise behind, add odd artifacts, or make a voice sound unnatural when the Noise Profile, the settings, or the source recording are difficult. Before giving up, it is worth checking a few common causes. If the result still is not usable, you can upload the original recording to Noisero and hear a short cleaned preview.

Drop your original recording here

or click to browse

We'll clean a short preview first. No payment required.

MP4, MOV, MP3, WAV

Tip: upload the untouched original recording rather than an already-processed Audacity export. Stacking noise-reduction passes on top of each other usually makes the audio worse, not better.

Hear the difference

Play the same moment with and without AI cleanup. The audio changes, but the video stays the same.

0:00 / 0:00

Why Audacity Noise Reduction may not work as expected

“Not working” can mean several different things. For some people the noise is still clearly audible; for others the voice now sounds robotic or thin, too much of the speech has been stripped away, or the whole recording picked up strange artifacts. These are separate symptoms with different causes, and naming yours makes it much easier to fix:

  • The Noise Profile does not represent the noise that is actually under the voice.
  • The section used for the profile accidentally contained speech or another wanted sound.
  • The noise changes throughout the recording, so a single profile cannot describe it.
  • The noise is almost as loud as the speech, leaving little room to separate them.
  • The noise overlaps heavily with speech frequencies, so reducing one affects the other.
  • The Noise Reduction (dB) amount or Sensitivity is set too aggressively and damages the voice.
  • The settings are too weak, so the noise is barely touched.
  • The source is clipped or distorted, so detail was already lost at recording time.
  • Several different noise sources are mixed into the same recording at once.

There is also an important design point. Audacity’s Noise Reduction is built around relatively constant background noise — steady hiss, hum, fan noise, whistle or buzz that stays much the same throughout. It is far less suited to irregular sound such as changing traffic, an audience, another person talking, music, or isolated clicks and pops. If your problem is one of those, no combination of settings will fully solve it, because it is not the kind of noise the effect was made for.

Check your Noise Profile first

This is the single most important thing to get right in Audacity. Before it can reduce anything, the effect needs a short sample of the unwanted noise on its own so it can estimate what to remove. You select a noise-only region, open Effect › Noise Removal and Repair › Noise Reduction , and click Get Noise Profile . Then you select the audio you actually want to clean and run the effect again with the Reduce option.

If the result is poor, the profile is often why. The most common mistakes are:

  • Selecting a section that also contains speech, breaths or other wanted sound.
  • Choosing a moment with a different kind of noise than the one under the voice.
  • Using an unrepresentative instant instead of the typical background.
  • Relying on one profile for a recording where the noise changes significantly.
  • Having no clean noise-only section at all — sometimes the recording simply never captured one.

A good profile should sound like the background you hear underneath the voice — the same hiss or hum, captured cleanly. Audacity only needs a short noise-only sample to build a profile, but a longer representative section is generally better when one is available. What matters most is that the sample contains only the unwanted noise and represents the same background sitting under the speech — that counts for more than choosing the shortest possible clip. If there is genuinely no noise-only gap anywhere in the recording, that is a strong sign the file itself, rather than the settings, is limiting the result.

If Noise Reduction seems to do almost nothing, the two-step workflow is often why. Clicking Get Noise Profile only samples the noise; it does not change the audio on its own. You then have to select the part you want to clean and open Noise Reduction again to apply it with the Reduce option. It is also worth checking the Noise Reduction (dB) amount: if it is set very low, close to zero, the audible effect will naturally be minimal.

What to try when Audacity Noise Reduction sounds bad

  1. 1

    Start again from the original recording. Undo previous attempts or reopen the untouched file. Stacking several failed noise-reduction passes on the same audio tends to compound artifacts rather than fix them.

  2. 2

    Create a better Noise Profile. Re-select a representative section that contains only the unwanted noise, then click Get Noise Profile again. A cleaner profile often does more than any slider.

  3. 3

    Use less aggressive settings. If the voice sounds damaged, lower the Noise Reduction (dB) amount and ease the Sensitivity. Reducing the noise a little less, but keeping a natural voice, usually sounds better overall.

  4. 4

    Listen to the Residue. Switch the Noise radio button from Reduce to Residue to hear what Audacity plans to take out. If you can recognise words or breaths in the residue, the settings are removing useful audio and should be eased back.

  5. 5

    Preview a short section first. Use Preview and experiment on a few seconds that contain both voice and noise before processing the whole recording, so you can judge changes quickly.

  6. 6

    Confirm the noise actually suits the effect. Steady hiss is very different from changing traffic, music, clipping or another person speaking. If it is one of the irregular cases, no setting will make Noise Reduction the right tool.

Why Audacity Noise Reduction can make voices sound robotic

This is one of the most common complaints, and it comes from a real limitation rather than a bug. Noise and speech frequently live in the same frequency regions. When the settings classify too much of the useful signal as noise, the effect carves out parts of the voice along with the background, and what remains can sound:

  • Robotic
  • Metallic
  • Watery
  • Hollow
  • Thin
  • Warbling or “underwater”

The key idea is that stronger processing is not automatically better. The controls that matter most here are Noise Reduction (dB) , Sensitivity and Frequency Smoothing . There is no single magic combination that works everywhere, because recordings differ — a value that rescues one clip will ruin another. Aim instead for the smallest amount of processing that makes the noise acceptable without noticeably harming the voice. When in doubt, reduce the amount rather than pushing every slider higher. If artifacts remain, experiment with Frequency Smoothing, but do not assume that increasing it will always improve the result — higher smoothing can also reduce clarity.

Audacity Noise Reduction works better on some noises than others

The effect models a steady background and subtracts it. That makes it a good fit for constant sound and a poor fit for anything that keeps changing:

Good candidates

  • Microphone hiss
  • Fan noise
  • Air conditioning
  • Steady hum
  • Stable background whine or buzz

Harder cases

  • Traffic that changes constantly
  • Keyboard clicks
  • Doors and impacts
  • Audience noise
  • Music or another person talking
  • Strong wind gusts, severe clipping or distortion

If your recording is dominated by one steady type, a focused guide can help — for example how to remove hiss from audio, how to remove hum from audio, or how to remove fan noise from audio.

Noise Reduction is not the same as repairing clipping or distortion

A lot of wasted effort comes from adjusting Noise Reduction settings for a problem it was never meant to solve. It helps to keep three different things apart:

  • Noise: unwanted sound recorded alongside the useful signal, such as a fan or steady hiss.
  • Clipping: the recording overloaded and parts of the waveform were lost at capture time.
  • Distortion: the source or recording chain altered the useful signal itself.

Noise Reduction can lower unwanted background sound. It cannot rebuild information that was never captured, so a badly clipped recording will not become clean no matter how the sliders are set. Recognising this early saves you from endlessly re-tuning the wrong tool.

Try the original recording in another noise remover

If you have already tried a representative Noise Profile and moderate settings but Audacity still leaves too much noise or damages the voice, the most useful next step is a direct comparison on the same source. Different tools analyse and separate sound in different ways, so the same recording can respond differently.

With Noisero you can:

  • Upload the untouched original recording straight from your device.
  • Hear a short cleaned preview for free before processing the whole file.
  • Compare the before and cleaned audio for your own recording.
  • Decide based on the actual result, not a general promise.

This is not a claim that one approach always wins — Audacity is a capable editor, and some recordings will do better there. The point is simply that comparing the same original source takes only a moment, and lets you choose whichever result actually sounds best for your file. If you find you mainly need the cleanup and not the editor, Noisero also works as an online Audacity noise reduction alternative.

How to avoid the problem in future recordings

Cleanup is always easier when the recording gives it good material to work with. A few habits make a real difference:

  • Record a short test and listen back before the real take.
  • Monitor with headphones so you catch noise as it happens.
  • Keep the microphone closer to the speaker.
  • Reduce obvious background noise before you start.
  • Set sensible gain and avoid clipping the input.
  • Capture a few seconds of room tone for a clean Noise Profile.
  • Avoid recording the voice very quietly and amplifying it heavily later.

Frequently asked questions

Why is Audacity Noise Reduction not removing all the background noise?+
Noise Reduction lowers unwanted sound rather than deleting it. When the noise shares the same frequency range as the voice, or is nearly as loud as the speech, some of it has to stay behind so the voice is not damaged. It also assumes a steady background it can model from your Noise Profile — if the noise keeps changing, one profile cannot describe it all.
Why does Audacity Noise Reduction make my voice sound robotic?+
Speech and noise often overlap in the same frequencies. When the settings treat too much of the useful signal as noise, the leftover audio can sound metallic, watery, thin or warbling. Lower the Noise Reduction (dB) amount and ease the Sensitivity until the voice sounds natural again — the strongest setting is rarely the best one. Frequency Smoothing can help with artifacts, but more of it is not automatically better and can reduce clarity.
What should I select for the Audacity Noise Profile?+
Pick a short stretch that contains only the unwanted noise and no speech — the same background hiss or hum you hear underneath the voice. Then run Effect > Noise Removal and Repair > Noise Reduction and click Get Noise Profile. If the sample includes words, breaths or a different noise, the profile misrepresents the real noise and the result suffers.
Why does Audacity Noise Reduction leave strange artifacts?+
Artifacts usually mean the processing is too aggressive or the Noise Profile does not match the noise. Switch the Noise radio button to Residue to hear what Audacity intends to remove — if you can pick out pieces of the voice in the residue, the settings are stripping useful audio and should be eased back.
Does Audacity Noise Reduction work on traffic or people talking?+
Not well. Audacity's documentation notes it is designed for constant background sound like hiss, hum, fan noise, whistle and buzz. Changing traffic, audience noise, another person talking, music and isolated clicks are irregular and do not fit the steady-noise model, so results are usually poor for those sources.
What can I try if Audacity Noise Reduction still is not good enough?+
Start again from the original recording, build a cleaner Noise Profile, use moderate settings, and preview a short section before processing everything. If the recording itself limits the result, upload the untouched original to Noisero and hear a free cleaned preview so you can compare the same source directly.