Export at a sane bitrate.
A 64 kbps MP3 has already thrown away most of what the denoiser would work with. If the recorder or the editor can give you 192 kbps or higher, take it. WAV costs you disk space and nothing else.
Clean Up Audio: What Noise Removal Can and Can't Fix
You already have the recording. That is the situation this page is written for. Somewhere on your machine sits a voice memo, an interview, a lecture capture, or a field take, and under the voice there is something you did not ask for. A hum. A hiss. A laptop fan that was six inches from the microphone. Wind. A room with bare walls that turned every sentence into a small echo.
NoiseClear applies background noise removal to that file in the cloud and hands back the same recording with less of that layer. There is nothing to install and nothing to configure. What follows is what the process removes, what it does not remove, and what the free tier includes, so you can tell in advance whether the file you have is worth running at all.
How to clean up audio on NoiseClear
Three steps.
Upload the file on the home page. Audio goes in as MP3, WAV, or M4A. Video works too, and that case is covered on the remove background noise from video free page.
Wait. A short file usually finishes in under a minute. A long one takes a few minutes.
Listen, then download.
The free tier adds no watermark, allows files up to 10 minutes long, and gives you ten minutes of audio a month, counted against your account rather than your browser. The 10-minute ceiling is the part people miss, so check the runtime before you upload.
Files and download links are deleted 60 minutes after upload, so pull your copy out of the browser while the link is live.
What clean up audio can remove: hum, hiss, wind, room echo, keyboard and fan noise
Denoising depends on the noise staying put. When it is steady, the model can learn its shape and subtract it. When it jumps around, you are asking for more than the method delivers, and the result starts to sound like it went through a filter.
What usually comes out cleanly:
- Electrical hum. The 50 or 60 Hz buzz from a grounding loop, a USB bus, or an interface sharing power with a monitor. Predictable, so it comes off well.
- Hiss. Tape hiss, the noise floor of a cheap preamp, or a quiet room recorded with the gain pushed up. Broadband and constant, which makes it the easiest item on this list. There is more on that in remove hiss from audio.
- Fan and air conditioner drone. A laptop fan, a window unit, a server left running in the closet.
- Keyboard and mouse clicks. Short transients get pulled down, not erased. Push the setting hard and you start biting into the consonants right after the click.
- Room echo. Short reflections from a room with parallel walls smear consonants and cost you intelligibility. Most of that comes back. The small-room case is covered in remove echo from audio.
- Wind on a phone mic. Low-frequency rumble from air moving across the capsule. The worst of it can be filtered.
Street noise and crowd noise are case by case. A steady bed of traffic or restaurant chatter can be pulled down several notches. A single motorcycle passing in the middle of a sentence is a peak, not a bed, and it stays.
What clean up audio cannot fix
This list matters more than the one above. It is the reason someone uploads a file, listens to the result, and decides that the tool does not work.
- Audio that is already clipped. If the recording was too loud and the peaks hit the ceiling, that information is gone, and the waveform is flat where it should have a shape. No denoiser puts it back. A restoration pass can make the crackle less harsh, but it cannot recover detail that was never stored.
- One instrument out of a finished mix. Pulling a vocal out of a mastered song is source separation, a different problem with different models. Noise reduction lowers the floor under the mix. It does not unmix it.
- Long reverb tails. A conference room is recoverable. A church or a stairwell puts hundreds of milliseconds of tail on every word, and the tail lands on top of the next word.
- Codec artifacts. The watery sound of a 64 kbps MP3 is damage from the encoder, not noise added on top of a signal. Denoising tends to smear it further. If that is all you have, go back to a better source.
The short version: if the noise was added on top of the signal, there is a good chance it comes off. If the signal itself was damaged, it stays damaged.
What changes and what does not
Denoising changes the noise, not the timeline. A six-minute recording comes back six minutes long, in the format and sample rate you sent. Nothing is trimmed, stretched or re-cropped, so a file that has to line up with picture still lines up after the run.
File size is set by the container and bitrate you chose, not by the cleanup. A WAV at 44.1 kHz holds the same number of bytes per second before and after, and a constant-bitrate MP3 pads to the same frame count either way. The recording sounds different and weighs about the same.
How far the noise drops depends on the noise. Steady hum, fan and air-conditioning tones sit in their own narrow band and come off cleanly. Irregular noise, a crowd or a busy street, is spread across the whole spectrum and can be reduced but not removed. A damaged signal is a different problem again: if the input is already clipped or distorted, the cleanup takes the background down and leaves the damage where it is.
There is no honest number for how much better a file will sound until someone listens to it. The test costs ten minutes and no money: run the file through the free allowance, keep both copies, and compare them on the same pair of headphones.
Formats, file size, and length limits
Audio in: MP3, WAV, M4A. Video in: MP4, MOV. The audio track is cleaned and put back on the original picture.
Ceiling per file2 GB and 60 minutes
Free tierup to 10 minutes per file, 10 minutes a month per account, free account, no watermark
Paid$1.00 up to 30 minutes · $1.99 30 to 60 minutes · or Pro at $9.99 a month
Deletion60 minutes after upload, for both the file and the link
Timingshort files usually under a minute, long files a few minutes
Getting a better export before you clean up audio
The file you upload sets the ceiling on the result, and a few habits raise it.
Keep the original sample rate.
Resampling 48 kHz down to 44.1 kHz before denoising moves the noise around and leaves odd artifacts near the top of the spectrum. Let the tool handle it.
Skip the normalizer and the noise gate.
Normalizing to 0 dB pushes quiet passages up until they clip, and most editors' noise gates chop the front off every word. Both make the file harder to clean than the raw take.
Do not stack passes.
Running three denoisers in a row does not triple the effect. It removes detail each time, and the third pass is usually audible as a hollow, metallic edge.
When you should clean up audio someone else recorded
A good share of the files people bring here were recorded by somebody else: a freelance interviewer's handheld, a client's phone, a lecture captured from the back row, a field take a colleague mailed over. Three habits save time.
Work from the original, not a re-export. Exporting from an editor and importing the result again adds a generation of lossy encoding, and each generation makes the denoiser's job harder. If the source is a video file, upload the video and let the tool pull the audio out.
Choose the cleanest copy that exists. If you have the recorder's WAV and an MP3 someone emailed around, use the WAV.
Do not expect a rescue. Room tone, hum, and fan noise come out. A blown-out recording does not, at any setting. Tell whoever is holding the mic next time to leave headroom and to point it away from the air vent.
Clean Up Audio: FAQ
No. The output runs the same length as the input. Nothing is trimmed and nothing is stretched, so an episode keeps its runtime and still lines up if you drop it back into a video edit.
Not the way you want. Once a peak has been clipped, the data at that peak is gone. Restoration can make the crackle less harsh, but it cannot rebuild the missing waveform. If every loud syllable crackles, the recording was made too hot and there is no way back from it.
No. That is source separation, not noise reduction. Upload a finished track and ask for the vocals alone and you get the track back with a lower noise floor, vocals still sitting inside the mix.
The free tier stops at 10 minutes. A file up to 30 minutes costs $1.00 to process. A file between 30 and 60 minutes costs $1.99. Each charge is for one file, not a plan.
Yes, a free one. You sign in with an email, and then the upload goes straight to processing. Paid runs use the same account, so paying adds nothing to set up.
60 minutes from upload, for the uploaded file and the download link both. Nothing is kept after that.
MP3, WAV, and M4A for audio. MP4 and MOV for video, where the audio is cleaned and put back on the original picture.
In the cloud, not on your machine. Nothing to install, and nothing running in the background once you close the tab.
Upload one file and hear it
Take the recording you have been meaning to fix. If it runs under 10 minutes, the free tier takes it under a free account, and without a watermark, and you will know inside a minute whether the noise comes off. Longer files have two one-time rates, both listed under clean up audio pricing. Anything this page does not answer goes to support@noiseclear.org.
Upload and remove noise