A recording with a fan or a fridge hum under it is hard to listen to, and the noise makes the voice harder to follow. The quickest fix is a browser tab, and several of them cost nothing. The best fix depends on how often you do this job and whether the recording can leave your machine.
Below are six tools that remove background noise for free, five in a browser and one installed app. For each one the list gives the exact free limit, where the processing happens, and whether the file is uploaded.
The free limits were read from each vendor’s own site on 16 November 2026. No tool here gave us a paid account, and no cleanup was run for this page: the numbers are the vendors’ published limits, not our own results.
Gleam Audio: free on your own computer
Gleam Audio is a desktop app for macOS, Windows and Linux. Remove noise takes out fans, hum, traffic and room noise under speech with a model that runs on your machine, so the recording is never uploaded. The same window then levels the volume, trims or splits the file, and converts it to the format you need.
The free tier is the whole app for personal use, up to 20 files a week, with no cut-down demo. Push a folder of files in and it cleans the list in one run, which is the part most free tiers charge for.
The trade-off is that it is an install, and it is built for speech. A music bed can lose detail, and there is no multitrack timeline. If a client recording is the job, the free tier does not cover it and the cheapest license is $10.
Voiser: free in the browser, no account
Voiser’s free tool takes a recording of up to 300MB, learns the noise from the quietest moments, and subtracts it. It offers light, medium and strong settings, with a before and after comparison over the first 10 seconds, which is the right way to choose a setting.
The part that matters most: the cleanup runs in the browser and the file never reaches their servers. Nothing to sign up for, nothing to install.
The method is classic spectral subtraction, and the page says so. Steady noise such as a fan or computer hum drops away. Noise that changes, such as a television or a crowd, cannot be separated from the voice, and echo is not noise, so it stays. On the strong setting voices can sound muffled. For a single private file, this is the best free option in the list.
Auphonic: two hours free each month
Auphonic is a cloud service built for spoken audio. It reduces noise and reverb, levels the speech, and measures the result against a loudness specification for a platform, which is the part none of the other free tools do properly.
The free tier processes 2 hours of audio a month. Productions made on that tier carry a short jingle, which rules it out for anything you plan to publish as it is.
The trade-off is the upload. The recording goes to their servers, then comes back. For a private interview or a client call that is a decision, not a detail. Our comparison of Gleam Audio and Auphonic covers where each one is the better choice.
Media.io: clean a video’s sound in the browser
Media.io’s noise reducer accepts audio and video files up to an hour long, so an mp4 or a mov can go in and come out with a cleaner soundtrack, without extracting the sound first. That is useful when the noisy recording is a clip rather than an audio file.
The free route needs an account: new users get free credits after signing up or logging in. The tool handles steady noise such as fans, hum, hiss and light wind, and the page is clear that it struggles when the noise overlaps the voice.
Media.io is a suite of AI tools, so the noise reducer shares a site with video and image generators. Expect the surrounding interface to be busier than the tool itself.
LALAL.AI: when music is under the voice
LALAL.AI is a separation service, and its Voice Cleaner tool takes out background music, plosives, mouth noise and microphone rumble. That is a different job from removing a fan: a music bed sitting under a voice is not noise, it is another signal, and subtracting a noise profile will not touch it.
The Starter plan is free and permanent, with 10 minutes of processing in the relaxed queue and a 200MB limit per file. The relaxed queue processes files as capacity allows, so it is not the plan to use when a deadline is close. Lite is $7.50 a month, billed yearly, for 90 minutes of fast queue processing.
The trade-off is the small free allowance. Ten minutes is enough to judge the output, not to clean a season of episodes.
CleanNoise: hear it free, pay only to download
CleanNoise is the honest middle ground. The 30 second preview is free at full quality, with no account, so you hear exactly what the paid file will sound like before you spend anything. A single download is $0.99 with no account at all.
It takes audio or video up to an hour, can run a second pass to take out echo and reverb, and deletes uploaded files within 48 hours. For one recording you will never touch again, that is cheaper than a subscription you forget to cancel.
The trade-off is that the file is not free, only the listen. There is no batch on the per-file route.
What each free plan actually limits
The word free covers six different arrangements, and the difference decides which one you can live with.
- Gleam Audio gives every tool and stops at 20 files a week, personal use only. Batch cleanup is included.
- Voiser gives the whole tool with no file count, but only for the job the page does: steady noise, one file at a time, in the browser.
- Auphonic gives 2 hours of processed audio a month, and free productions carry a jingle.
- Media.io gives free credits to new accounts, then a subscription.
- LALAL.AI gives 10 minutes in the slow queue, once, as a Starter allowance.
- CleanNoise gives a 30 second listen at full quality, and charges $0.99 per download.
If the recording is a one-off, CleanNoise or Voiser covers it at no cost. If you clean audio every week, the file count matters more than the price, and that is where an installed app with an uncapped batch starts to win.
Where a browser tool stops working
Two problems sit outside what any of these free tiers fix.
The first is anything mixed under the voice. A television, a crowd or a music bed is not steady noise, and a noise profile cannot subtract it. LALAL.AI’s separation approach is the exception, and it is why that tool is in the list.
The second is a voice that has already been damaged. Aggressive noise reduction leaves artefacts, and a voice can end up sounding thin or robotic. Our article on why noise reduction makes a voice sound robotic covers the causes: too much reduction, a noise profile taken from a section that already contained speech, and repeated passes over the same file.
The recording itself decides how much cleanup you need. A microphone a hand’s width from the mouth, in a room with soft furnishings, gives every tool in this list far less to do.
Which one fits your job
If the file is a single private recording and it must not leave your machine, use Voiser, or install Gleam Audio and clean it offline.
If you clean speech every week, install Gleam Audio. The free tier covers a batch, and the Remove noise tool is the one to start with.
If the file has to hit a podcast or broadcast loudness target, use Auphonic or finish with Gleam Audio’s Enhance tool and check the number against the LUFS guide.
If the recording is a video clip, try Media.io, or use Gleam Audio’s Remove noise on the file, which saves the sound only.
If music sits under the voice, use LALAL.AI.
If you want to hear the result before paying anything, use CleanNoise’s preview.
For a quick check with nothing installed, the free browser noise remover on this site runs in the tab, and remove noise from video online covers the video version of the same job. Adobe Podcast Enhance is the other free browser tool people start with, and our page on Adobe Podcast Enhance alternatives covers what happens when its free processing is not enough.