Improving noise reduction

I know it’s my recording device that picks up a ton of noise, but is there a way for Audacity to improve the noise reduction? I’ve noticed, that the more you adjust the slider to get rid of more db noise, that it makes things tin like, and makes it worse quality than before. For another, noise reduction does not pick up all the noise I’d like it too. Sometimes I have to run noise reduction over the same spot 3 to 5 times, and that’s adjusting the sliders to a reasonable spot. For me, noise reduction works, and it’s a great tool, but it’s almost a game some days just to find out how much db I want to remove, without making it tin like, and how little sensitivity I have to use in order for it to pick up just a tiny bit of the noise.

So what I’m asking, is if there is a way that noise reduction can pick up the noise better?

So what I’m asking, is if there is a way that noise reduction can pick up the noise better?

You probably want Audacity to reduce more noise better. Audacity picks it up just fine.

This is the improved Noise Reduction. The older version only reduced noise in the spaces between sounds or voices. If you had enough noise, it would give you pumping noise with voice volume. Not the best special effect.

I know it’s my recording device that picks up a ton of noise

What’s the job? You didn’t say. It’s good not to address issues in a vacuum.

I published a Kitchen Table Studio which helps get rid of environment noises and echoes.

Depending on the job, that may be enough.

If most of the noise is microphone hiss (ffffff), that may be the end of the world. The only way out may be to change the microphone.

There is also Noise Gate.

The service is tuned to recognize the spaces between words or sounds and only reduce noise during those spaces. That gets rid of the tinny voice problem, but it’s terrifically hard to adjust and keep working because it can’t tell when you announce louder or softer, and it has odd sound problems of its own.

You would think very high technology should be able to pull this out, right? They did. That’s how cellphones work. On the fly, adaptive noise reduction. Unfortunately, the goal there was cellphone voice, not theatrical presentation.

Is that your goal? You didn’t say.


Yes, I meant to reduce noise, but it also doesn’t “pick” up noise well enough either IMHO, because sometimes I can select a very small amount as the noise profile, and it still, no matter what I select for the sensitivity or how much db of the noise I want to remove, it doesn’t remove all the noise. There’s still a faint amount that you can hear good enough.

Well that’s good news it’s improved immensely over the years, hopefully it can improve on the small problems I’m experiencing.

I know it’s my recording device that picks up a ton of noise

It collects a good amount of noise, I don’t use it anymore, but I was using that recording device as an example. The point of this post, was to ask if there was a way that noise reduction could be improved for those devices which we love, but pick a lot of noise up, and make it difficult to remove.

Goal? The goal is to see if audacity can improve the noise reduction. I’m not having problems with it now that I change device.

I’m not having problems with it now that I change device.

What did you go from and to? This is a forum with users helping each other. It would be good to know that. We have been known to recommend kit or not according to users experiences.

Being able to shoot whatever you want and turn it into a marketable product with post production is a terrific goal. We do have advances like that occasionally. There’s a new tool intended to be useful for audiobook readers for one example.

Thanks for your post.


The device that I changed too, was simply recording straight from Audacity, and I’m not looking for post production or marketing products! I am trying to help out members and I understand what this forum is for. All I was asking for was if Audacity could improve the noise reduction, and so I shared my experience by testing some things out to write up some feedback on how the noise reduction did, and my thoughts on it… how did that come across as wanting money and making that my goal of being here? If I’m posting some how to’s, and trying to give some hints, doesn’t that show my trying to help a little? Probably not because I’m new, and I really haven’t been here long, so you don’t know if you can trust me yet; I get it, and I apologize for coming across as a newbie wanting attention.

you don’t know if you can trust me yet

No, and you’re not making it any better. I don’t trust posts with no back story.

There is a dark side to Audacity. One of the common requests is for improved noise reduction to help dig voices or sounds out of heavy noise and interference for the purpose of surveillance, conflict resolution and law enforcement.

Audacity can’t be used for any of those things.


Oh, I didn’t know that. I was seriously just requesting this to help improve my podcasts and music… now I’ve made a fool of myself :blush:

I do have a question. I went to audacity earlier today, and I when I used the noise reduction, I selected a small, odd noise that had appeared three times in the recording, but audacity gave me a pop-up saying “selected file too short”. There wasn’t much silence between the words that the noise was embedded in, so I really couldn’t do much about it. Instead I had to cut it, and paste silence to make it sound natural. But one thing that could be improved on noise reduction, is being able to select smaller spots. That way I would make a one step process for those tiny odd noises (kitty cat meows).

A little “trick” you can do in such cases, is to temporarily repeat the selected noise a few times (Effect menu > Repeat) so that it is long enough for the Noise Reduction effect. Then take the Noise profile. Then Undo do repeats (Ctrl + Z). Then you’re ready to apply Noise Reduction.

Noise reduction effects have to do a delicate balancing act between removing what appears to be noise, and retaining what appears to be “not noise”. Audacity’s “Noise Reduction” effect does this pretty well for low level constant noise (such as tape hiss), though there’s probably room for a bit of improvement if any skilled C++ developers are interested in having a go.

Thank you Steve. I tried it, and this helped a bit. Certainly improved… wonder how many people listening to my podcasts will notice the difference :wink:

I think this is the basis of your problem. A kitty cat meow is not “noise”.

The noise reduction tool is designed to remove actual noise, like a constant hissing from the microphone itself, or a constant unwanted noise from the environment (such as a computer fan), or even something like tape hiss if you’re transferring audio from tape. When you’re recording you should always record a few seconds of “silence”—no talking, breathing or kitty cat sounds—and then use those few seconds of “silence” as the noise reduction tool’s noise profile.

This is possible, but almost irrational. That is, select a plot with noise, then repeat, then return it back, undo. The effect is already of two steps, and so will all 4.I wanted to suggest another improvement for the noise reduction effect. That is, after step 1, the noise model, a reminder appears. Because I often then just forget to apply it. Everything is written here:

I’m going to disagree with you. My mic picks up almost everything, and that kitty cat noise is certainly a “noise”. A sound is a noise of some sort.

That’s little bit hard to do, when the noises are inside the track. I also understand getting rid of the hissing part.

Hmm… I didn’t notice this before. It certainly helps. Thanks for the advice!

OK, I’ll grant that a meow sound is “noise” (it’s part of the “environmental noise”).

But the fact remains that Audacity’s Noise Reduction tool is designed for and works best at reducing the random “hiss” type of noise that is produced by audio equipment (mic, pre-amp, DAC, analog media, etc). I suppose it can be used to reduce some environmental noises like meows, but I imagine it wouldn’t be very effective.

I suppose it can be used to reduce some environmental noises like meows, but I imagine it wouldn’t be very effective.

No, but maybe not the the reason you think. Noise Reduction doesn’t take into account perceived loudness.

If you have relatively well-behaved microphone noise (gentle spring rain in the trees—ffffffffff) then a modest application of noise reduction can work very well. This is the Noise Reduction of the Beast 6, 6, 6. However if you have The Yeti Curse USB microphone noise, it’s going to work a lot less well—even at identical volume. Yeti Curse has a lot of the characteristics of Baby Screaming on a Jet. Just try and ignore that for a six hour flight. It works so badly that we designed a tool specifically to deal with it.

Noise Reduction has a separate training step for a reason.

I agree that kitten noises should be a standard setting for Noise Reduction. Which kitten shall we use?

I’ll wait.


There is a way to do this automatically. The way cellphones and conferencing system do it. They assume any tones that stick around for longer than a set time are probably noise. FanNoiseFanNoiseFanNoiseFanNoiseFanNoiseFanNoise

In their application, they do very well. The problem is using them outside their application. Try to get music past one of those.

“How come my music keeps fading out?”

There is an additional problem with audiobooks. ACX Hates Noise Reduction and they say so. ACX tests your submissions for volume, overload and noise semi-automatically as your submission arrives. But then it goes on to Human Quality Control, and if the voice doesn’t sound clear, perfect and natural, you lose.

Very few heavy noise reduction jobs sound clear, perfect and natural, no matter how they’re done.

I’d be happy if Audacity Noise Reduction didn’t forget my profile until I put a new one in. Or better still, also let me save a profile without all that sound clip management.


In my experience, noise reduction works well for most things, including kitty noises, however, there are certain noises which have a few problems. And yes, environmental noises can we removed simply through noise reduction. I’ve done it, I’ve seen it happen, and I know others who are able to do it. You can always of course, re-do the part in which the “environmental noise” comes in, but either way, I’d rather have the noise reduction take it out, rather than re-do it, and probably have it pick up at maybe a higher or lower pitch than the moment before and after. When reading for say, 20 minutes, your voice get’s weaker, and maybe higher pitched, so jumping at the 20 minute mark with a “new” voice, is going to make things sound oddly.

Nice how you give those noises their proper titles, but they aren’t anything anymore since noise reduction came along. Same applies with the endless Kitty cat meows