How to Remove Background Noise From a Recording
A recording with a hum, a fan, street noise, or general hiss underneath the voice is one of the most common audio problems, and also one of the most improvable with free tools, as long as you're realistic about what noise reduction can and can't fix. Here's how to actually do it, and what to expect from the result.
How Do You Remove Background Noise From a Recording?
In Audacity, select a short section containing only the background noise (no voice), use Effect > Noise Reduction > Get Noise Profile to learn it, then select the whole track and apply Noise Reduction again to reduce that noise throughout. This handles steady noise, hums, fans, hiss, well; it's not built for removing background music mixed with a voice.
Setting realistic expectations matters here: noise reduction improves a recording meaningfully, it rarely makes it sound like it was recorded in a professional studio. Understanding the technique, and its limits, gets you the best free result.
The Core Technique
- Open your recording in Audacity.
- Find a clean noise sample, a short stretch (2-3 seconds is enough) where only the background noise is present and no one is talking, often right at the start of a recording before anyone speaks.
- Select just that noise-only section by clicking and dragging over it in the waveform.
- Go to Effect > Noise Reduction and click "Get Noise Profile." This teaches Audacity what the background noise sounds like, without applying anything yet.
- Select the Edit menu and choose Select All (or Ctrl+A) to select the entire track.
- Open Effect > Noise Reduction again, and this time adjust the settings before clicking OK to apply the change.
- Listen back to the result, and undo and readjust if the voice sounds distorted or watery.
According to Audacity's own documentation, this two-step process, learning the noise profile first, then applying reduction to the whole track, is the core of how the tool works, and skipping the profile step or using a bad sample (one that accidentally includes some voice) is the most common source of poor results.
Avoiding the Robotic Sound
Audacity's Noise Reduction effect has three main settings, and getting them wrong is what produces that unpleasant robotic or watery artifact people associate with over-processed audio.
| Setting | What it controls | Common mistake |
|---|---|---|
| Noise reduction (dB) | How much the identified noise is reduced | Set too high, causing artifacts |
| Sensitivity | How aggressively the tool identifies noise vs. voice | Set too high, cutting into the voice itself |
| Frequency smoothing | How the reduction is spread across frequencies | Set too low, causing a "musical" warbling artifact |
A commonly recommended starting point is a noise reduction value in the 6-12 dB range rather than pushing to the maximum, since more aggressive settings are where the artifacts become noticeable. Apply a moderate amount, listen critically, and only increase it if the noise is still clearly audible and the voice still sounds natural.
Why This Technique Doesn't Remove Background Music
It's worth being direct about a limitation: Audacity's Noise Reduction effect is built for steady, consistent noise, the kind of sound that has a stable, learnable "fingerprint" like a fan hum or air conditioning hiss. Background music is structurally different: it varies constantly in pitch and volume in ways that overlap with the frequencies of a human voice, which makes it a much harder separation problem that standard noise reduction handles poorly.
If your actual problem is removing background music from a video while keeping the voice, that's a different technique entirely, covered in our dedicated guide on removing background music from a video, which uses different tools built specifically for that harder separation task.
When Audacity's Approach Falls Short
For recordings with more complex or varying background noise, several newer, AI-based noise reduction tools (some free, some with free tiers) use machine learning models trained on many examples of noise and speech, rather than learning only from a sample of your specific recording. These can sometimes handle trickier situations, like noise that changes over the course of a recording, better than Audacity's more traditional approach.
For simple, steady background noise, Audacity's free built-in tool is often just as effective and doesn't require finding, installing or trusting an additional third-party tool.
The Best Fix Happens Before Recording
Noise reduction is a repair technique, and repair always has limits that prevention doesn't. If you're recording again in the future, a few habits reduce how much repair work you'll need afterward:
- Record in a quieter room, away from fans, open windows, or humming appliances.
- Use a dedicated microphone rather than a laptop or phone's built-in mic, which tends to pick up more ambient noise.
- Get physically closer to the microphone so your voice is recorded at a higher relative volume compared to the room noise.
- Record a few seconds of silence at the start deliberately, giving yourself a clean noise sample to work with later even if you don't end up needing it.
What Trips People Up
- Using a noise sample that includes some voice. This confuses the profile and can cause the tool to reduce the voice itself. Find a genuinely voice-free section.
- Pushing the reduction setting too high. As covered above, this is the main cause of the robotic, watery artifact; moderate settings applied thoughtfully beat maximum settings applied carelessly.
- Expecting noise reduction to fix background music. As covered above, that's a different problem requiring different tools.
- Not listening critically to the result. Always play back the processed audio, ideally with headphones, before considering the job done.
- Applying noise reduction multiple times in a row. Repeated passes compound artifacts rather than removing more noise; if one pass doesn't sufficiently reduce the noise, adjust the settings and start over from the original rather than stacking passes.
What Comes Before or After
If your noisy recording is actually the audio track of a video, extracting the audio first lets you clean it up in Audacity before recombining it, or before exporting just the cleaned audio on its own. And if you're recording specifically to avoid capturing background noise from your microphone in the first place, recording your screen without the microphone covers a related but different scenario.
Frequently Asked Questions
What's the best free tool for removing background noise from audio?
Audacity's built-in Noise Reduction effect is the standard free option, working by learning a short sample of just the background noise and then reducing that specific noise profile throughout the rest of the recording.
Can background noise removal completely eliminate noise with no side effects?
Not entirely. Noise reduction reduces steady, consistent background noise well, a fan, a hum, air conditioning, but pushing it too aggressively introduces its own artifacts, a watery or robotic distortion. It works best as a meaningful improvement, not a perfect fix.
Does noise reduction work on background music, not just hiss or hum?
Not well. Removing background music mixed with a voice is a fundamentally different and harder problem than removing steady ambient noise, and needs a different approach, covered in our dedicated guide on removing background music from a video.
How do I get a clean noise sample for Audacity to learn from?
Find a short section of your recording, ideally 2-3 seconds, where only the background noise is present and no one is speaking, often at the very start or a pause in the recording, and select just that section before running Get Noise Profile.
Are AI-based noise removal tools better than Audacity's built-in effect?
AI-based tools can perform better on complex or varying noise, since they're trained on many examples rather than learning from just your specific sample, though results vary by tool and recording. For steady, simple background noise, Audacity's classic approach often works just as well for free.
More free software picks
We track free-tier limits and free tools across dozens of categories since they change more often than you'd think.
More from the creator
Explore two other projects from the same creator.