Noise Removal (and other effects) Advice

Hello - I’m working with a (so far) unavoidable source of background noise. Wondering if I could get some advice on the best way to remove it. I’ve been playing around with the Noise Removal settings (and other effect settings), and have reached the limits of what I can do. Wanted to see if any one else can find a better way to remove it.

I’m shooting for trying to hit the ACX submission requirements of RMS between -23db and -18db, have nothing higher than peaks of -3db, and a maximum -60db noise floor. These are arbitrary goals, not something I’m actually required to meet. Just figured that if I can hit those goals then generally I have pretty good sounding audio.

I’ve attached two examples. “Test.wav” is my un-edited recording. “Text_edited.wav” is the best my feeble brain can do to improve it.

Here was my process:

  1. Amplify the vocals only, setting the New Peak Amplitude to -.3db.
  2. Compress the vocals only
    Threshhold = -8db
    Noise Floor= -35db
    Ratio= 2:1
    Attack! = 0.2
    Release = 1.0
    Make-up gain not checked
  3. Noise Removal profile sample of the first two seconds
  4. Noise Removal
    Noise Reduction = 24
    Sensitivity = 13.82
    Frequency smoothing = 670
    Attack!/decay time = 0.07

Any takers to educate me on how I can improve things?

Thanks, any help or suggestions is much appreciated.

Can you wait a few weeks until the new year? I’ve been testing the alpha version of the new Noise Reduction effect which I’ve slightly modified, and in my tests it is producing much better results that the Noise Removal effect that is currently available.

Noise reduction of 24 in the current effect is probably too drastic and is likely to cause damage (metallic bubbly sounds) in the processed audio.

I didn’t see any samples posted.
Hearing the raw, unprocessed clip is a Really Big Deal. We can tell a lot very quickly from that.

Apparently I’m an idiot and forgot to post the samples. Here they are.


I’m in no rush, so waiting a few weeks is not a problem.


I can see one problem immediately. What did you use as the Noise Profile? The area before and after your sound sample isn’t the best. You were rumbly, thumpy doing stuff back there; you didn’t just stop talking. That can be deadly because it confuses Noise Removal. To Perform for Profile, hold your breath and stop moving around. Two seconds should do it, you don’t have to turn blue.

If you have activities in the background such as cars starting up, TVs or conversations, you need to fix that. That condition is one of the ways to reliably kill your show. There are no good tools to fix that.

I think you overdid the patched clip. The object is gentle removal, not dead quiet backgrounds. Overdoing Noise Removal can damage the voice in the show. Start with gentle values of removal such as 9, 12 or 18. Noise Removal doesn’t just attack noise. It attacks everything and it’s up to you to pick the trade-off place.

You may not need to wait for Steve. I think you can do it with the work you have here.


Thanks Koz. I was trying to hit the maximum -60db noise floor, and it didn’t seem like lower values of noise removal were getting me there. Maybe that’s because I had a pretty poor Noise Profile?

I got noise values in the 70s, and again, if you’re measuring a live, active room, that’s not a Noise Removal problem. Noise Removal only works on periodic, extended sounds like fans or hum and buzz. It doesn’t remove “Stuff I Don’t Like.” It has significant problems with hiss. Hiss is made up of all frequencies, so Noise Removal tries to remove the whole show.

There was also a push to rename it “Noise Reduction.” Too many people are expecting it to remove noise.

Did you actually have ACX reject your posting?


That brings up another discussion point. How are you measuring your work?

Nope, was just using their guidelines as a goal, figured if I could meet their guidelines then I’d have fairly good audio quality.

After running through my effects, including Noise Removal, I highlight a section that is just background noise, then go to Analyze > Contrast, and click a Measure Selection button. Maybe that’s not the correct way?


That’s one way. We get immediately into the different ways to measure noise. If you’re in the US, you are no doubt familiar with OSHA which puts those orange and white warnings on the ladders [You could fall and terminate your family tree]. They measure noise A weight which can give a significantly different reading than C weighing or Z. I think the Audacity tools measures Z. I could be corrected.

A takes into account the difference between pleasant distant thunder and Infant Screaming on a Jet.

There is a USB Microphone digital failure which illustrates this beautifully. I call it Frying Mosquitoes and it’s not very loud, but it’s right at that pitch that makes homicide attractive.

There are differing opinions on what ACX uses.


Not to go all crazy on you, but if ACX compliance was one single, simple thing, it would make everyone’s life a lot easier.

Use OSHA standards. Does extended exposure your show contribute to damage in unborn children?


You can’t prove anything!!!

Lol :smiley:

OK, back on topic…

Not by a lot, but yes I agree, a bit overcooked.

You’ve gone way below it, but I can see why you think you haven’t.

That’s going to give you a figure that is a lot higher than it really is, because of those thumps and clicks. If you miss the thumps and clicks, the rest measures at about -86 dB.

Ok, thanks, good to know.