In this guide
Select a passage containing only the same steady noise, capture its profile, then apply a restrained reduction to the wanted recording. Preview the removed sound as well as the result; audible words in the removed sound are a reason to adjust or undo.
To remove steady background noise in Audacity, select a section containing only that noise and choose Get Noise Profile in the Noise Reduction effect. Then select the recording to process, reopen the effect, preview a modest reduction and compare it with the original. Listen to the removed sound too: if it contains recognizable words or notes, the settings may be removing wanted audio.
This method works best when the noise is relatively consistent, such as a fan or hiss, and a representative noise-only section exists. A passing car, audience or individual click changes over time and needs a different approach.
Before you start: find a useful noise sample
Save an unchanged copy of the source. Listen to a passage with speech or music and find a pause where the same background is audible on its own. Avoid breathing, quiet words, music tails and clicks in the profile: the effect may treat them as noise too.
According to Audacity's Noise Reduction manual, a profile requires at least 2,048 samples—about 0.05 seconds at 44.1 kHz—and a longer representative sample is preferable. Repeating the same tiny sample does not create a better representation of the changing noise. If the background varies across takes, a profile from one scene may not suit another.
Step by step: reduce steady noise
- Open the recording and keep a reference. Work on a copy or duplicated track. Mark a representative passage containing a pause, quiet speech and the most noticeable background.
- Select only the background noise. Drag over the noise-only pause. Play it to ensure it does not contain a word, breath or instrument you need.
- Capture the profile. Open Effect → Noise Removal and Repair → Noise Reduction and choose Get Noise Profile. Menu labels can differ by version. The profile tells the effect what the selected noise looks like; it does not process the whole recording yet.
- Select the audio to clean. Select the test passage or the intended track, reopen Noise Reduction and preview a restrained setting. Listen to a complete phrase at a similar playback level to the original.
- Check what would be removed. Use the effect's removed-sound option—called Residue in the legacy manual and Noise only in the current manual. If you hear identifiable speech or music, try lower reduction or sensitivity, or choose a cleaner profile.
- Apply the wanted output. Switch back to the reduced-audio option, apply, and compare the full passage with the untouched copy. Undo if the speaker sounds metallic or words have changed; then check the exported file.
Do not delete the pause to make it quieter if your recording must stay in sync with video. A reduction effect changes the selected sound; deleting time moves everything after the cut.
What do Noise Reduction, Sensitivity and Frequency Smoothing do?
| Control | What it changes | Tradeoff to listen for |
|---|---|---|
| Noise Reduction (dB) | Amount of reduction applied to detected noise. | More reduction can make wanted sound less natural. |
| Sensitivity | How much audio is classified as noise. | More sensitivity may remove speech or music; less may leave artifacts. |
| Frequency Smoothing | Spreads reduction across neighboring frequency bands. | It can make artifacts less conspicuous but also soften wanted detail. |
There is no universal setting for every microphone or room. The Audacity manual recommends using the lowest reduction that makes the noise acceptable, and notes that smoothing can make desired audio less clear. Use the preview and the removed-sound check rather than copying settings from an unrelated recording.
Troubleshoot common results
| Result | What to try |
|---|---|
| Voice sounds metallic or watery | Undo, use a less aggressive reduction, and check whether the profile matches the noise during speech. |
| Words appear in Noise only or Residue | Lower sensitivity or reduction and check that the profile contains no speech. |
| Fan changes speed midway through the take | Compare sections separately and find a representative profile for each; preserve smooth transitions. |
| The profile button gives an error | Check that the selection contains enough samples and that profile and target sample rates match. |
| Noise disappears in pauses but remains under words | Listen at the overlap. A gate affects level in pauses; it does not identify or separate an overlapping sound. |
The legacy manual notes that applying Noise Reduction to time-stretched clips can render the change, and documents a potential interaction with volume envelopes. Keep a project copy and check edits and gain automation after applying an effect. Interface behavior can change by version.
What if there is no noise-only section?
Do not create a profile from a full sentence containing the voice you want to preserve. Compare another take or microphone recording if available. For a distinct competing source, you can test source isolation on a short excerpt: open the voice isolator, upload a supported MP3, WAV, MP4 or MOV up to 50 MiB (52.43 MB), describe “spoken voice,” and review Isolated sound. A separator can also alter words or retain noise; compare with the original before choosing it. The download dialog shows the cost and AI downloads require an account and sufficient credits.
For clicks, traffic, music or overlapping speakers, use the cleanup method comparison to select a starting approach. If the audio returns to a video, preserve its placement and follow the audio replacement checklist.
Common questions
Can Audacity remove background noise without losing voice?
It can reduce some noise while preserving useful speech, but no setting guarantees an unchanged voice. Compare a complete sentence and listen to the removed sound before deciding.
Can a noise profile remove music or another speaker?
Not reliably. The effect is designed for relatively constant noise; changing music and voices can overlap the wanted speech. Try original separate tracks or compare a source-isolation workflow.
Does Audacity keep the profile for my next session?
The legacy manual says the profile is not retained after Audacity closes. Capture a suitable profile again in a new session and verify the current version's behavior.