How to remove loud background noise from a job site video

This page may contain affiliate links from which Blur Studio earns commissions.

Job site audio is one of the harder cleanup jobs because the noise usually isn't one steady, predictable sound. It's a mix: machinery, traffic, wind, other crew talking in the background, all layered and changing throughout the clip. That's exactly the case where a basic noise-reduction filter struggles and an AI-based tool does noticeably better. If the problem is a voice agent that cannot hear callers, see how to make a voice agent hear well on a noisy job site.

Try the free option first, if the noise is simple

If your background noise is one consistent, steady sound, a hum, a fan, a constant low rumble, a free tool like Audacity's noise reduction filter can often handle it well enough, at no cost. It works by learning the noise profile from a quiet section of your clip and subtracting that profile from the whole recording. It struggles badly, though, once the noise is inconsistent or changes character throughout the clip, which describes most real job site audio.

Using an AI isolator for layered, changing noise

For the messier, real-world case, ElevenLabs' Voice Isolator works differently: rather than learning and subtracting one noise profile, it identifies speech directly and extracts it from everything else in the recording, which handles inconsistent, layered background sound much better than a filter built around one noise profile.

1. Upload your file to Voice Isolator. It accepts both audio and a wide range of video formats directly, so you don't need to pull the audio track out yourself first. The limit is 500MB per file and 1 hour of duration.
2. Let it process. This runs on credits, roughly 1,000 credits per minute of audio, so a 3-minute job site clip costs around 3,000 credits, a small fraction of what's included on most paid plans.
3. Check the result against the original, not in isolation. Isolation can occasionally soften or thin out the voice itself along with removing the noise; listen for whether the speech still sounds natural, not just whether the noise is gone.
4. Re-sync to your video if you pulled just the audio track out, or drop in the processed file directly if you uploaded video.

A worked example: a plumbing walkthrough with a jackhammer next door

Say you're filming a before-and-after walkthrough for a customer, explaining the work you did, while a crew on the next lot is running a jackhammer intermittently in the background. This is a genuinely hard case for a basic filter, since the noise isn't constant, it comes and goes, which means there's no single steady profile to learn and subtract. Here's the actual sequence:

1. Don't try to time your talking around the jackhammer. Record the walkthrough naturally; trying to pause and restart every time the noise kicks in usually produces a worse, more disjointed final video than just recording through it and cleaning up afterward.
2. Upload the full clip to Voice Isolator rather than trying to isolate just the noisy sections. Processing the whole file is simpler and the tool handles intermittent noise fine, since it's identifying your speech throughout rather than learning one noise pattern to subtract everywhere. The same cleanup helps before transcription, see how to clean up audio before transcribing it.
3. Listen to the loudest jackhammer moments first when checking the result, not the quiet sections. If the voice holds up clearly through the worst noise spikes, the rest of the clip will be fine too.
4. Keep the original file. If a specific few seconds still sound rough, it's sometimes faster to re-record just that short section separately and swap it in, rather than reprocessing the whole clip trying to chase one bad moment. Phone recordings can be rescued too, see how to make hard-to-hear recorded customer calls clearer.

The result is a walkthrough where the customer hears you clearly explaining the work, with the jackhammer reduced to a faint, inoffensive background texture instead of something that competes with your voice.

Free filter, AI isolator, or reshoot: picking the right one

Situation	Best option
Steady hum, fan or low rumble throughout	Free filter (Audacity's noise reduction)
Intermittent or layered noise (traffic, machinery, crosstalk)	AI isolator
Music mixed in with speech	Neither handles this reliably; separate audio and video capture, or a dedicated stem-separation tool, is the better route
Voice itself was poorly captured (too far, mumbled, clipped)	Reshoot; no cleanup tool recovers clarity that was never recorded
Predictable noise source (generator, compressor)	Prevent it on site (ask for a pause) rather than cleaning it up after

The table's real point is that "clean up the audio" isn't always the right answer. Matching the actual cause to the right fix saves you from running a clip through a tool that was never going to solve that specific problem. Echo from a remote call is a different cause, covered in how to fix echo and bad sound in a remote interview recording.

How to judge the result honestly, not just "is the noise gone"

The easy mistake is listening only for whether the background noise disappeared and calling it done. Listen specifically for two other things too: whether your voice itself still sounds natural (isolation can occasionally thin out or slightly flatten a voice along with removing the noise around it), and whether any processing artifacts, a faint warble or watery quality, appear in moments where the original noise was loudest. If either shows up, it's usually more noticeable on headphones than on a phone speaker, so do your final check on headphones before publishing anything client-facing. A/B it against the untouched original rather than judging the cleaned version in isolation; your ear adjusts to small artifacts quickly when you're not comparing directly against the source. If you're processing several clips from the same shoot, spot-check at least the worst one and the mildest one rather than assuming every file came out equally clean, since noise severity varies take to take even on the same job site.

What it won't fix

An isolator cleans up noise that's genuinely separate from the voice, it doesn't fix a voice that was badly recorded to begin with. If someone was speaking too far from the microphone, mumbling, or their voice was clipped or distorted at the source, no amount of noise removal recovers clarity that was never captured. It also generally does better removing ambient noise and chatter than it does cleanly separating a voice from music playing at the same time; if your clip has background music layered with speech, don't expect the same clean result you'd get from pure environmental noise.

For job sites specifically, wind noise hitting the microphone directly (as opposed to wind sound in the environment) is one of the harder cases, since that noise is often picked up as distortion in the recording itself rather than as a separate, identifiable background sound. If that's your specific problem, a cheap foam windscreen on future shoots will do more for you than any cleanup tool after the fact.

Preventing the problem on the next shoot, not just fixing this one

Cleanup tools are genuinely good, but a few minutes of prevention on site beats the best cleanup afterward. A cheap foam or furry windscreen on the microphone cuts wind-generated distortion at the source, which, as mentioned above, is one of the cases cleanup struggles with the most. Holding the phone or recorder a bit closer to whoever's talking, even a foot closer, dramatically improves the ratio of voice to background noise before any processing happens. And if a specific noise source is predictable, a generator, a compressor cycling on and off, asking whoever's running it for a short pause during the one minute you actually need clean audio costs nothing and saves a cleanup step entirely. None of this means skip the isolator; it means the isolator has an easier job, and an easier job means a cleaner result, every time, with fewer takes needing a second pass or a manual fix afterward.

How I'd approach this

For an occasional clip, running it through the isolator yourself takes a few minutes and costs very little. If your business is producing a lot of job site video regularly, training content, before-and-after walkthroughs, customer-facing updates, I'd build this cleanup step directly into whatever process already handles your video, so it happens automatically on export rather than someone remembering to run every clip through a separate tool by hand. For a business shooting several clips a week across different crews, that automation also means someone isn't quietly skipping the cleanup step on the days they're busy, which is usually how an inconsistent library of video quality happens in the first place. The same pipeline can also flag a clip for manual review when the processed result still sounds rough, rather than publishing it automatically and finding out from a customer.