Trying to fix a horrifically bad music recording

I am new to audacity and I am using version 2.0.1. My OS is Windows XP Home Edition Version 2002, Service Pack 3.

Problem: I have spent all day trying to fix a horrific recording from Jackson C. Frank called ‘In the Pines’(Youtube link: I assume that no known good copies exist of this song, as this very poor recording is part of his 42 song collection that is being sold on Amazon. I very much like this song and hoped that I could fix it using the noise reduction features in audacity.

The song only has a very short ‘noise only’ portion at the front end to work with for a noise profile, so to get a better profile I have tried amplifying it, de-amplifying it, copying it over and over and then selecting the entire compiled noise section (with/without amplification), and I cannot get rid of the noise in the song without destroying it (very muted). I thought to select the front end, and then copy some additional real tiny snippets of noise from within the song to stitch the noise into a single section that I can then select for the noise profile. I don’t know if anyone has tried that. Since you folks are the experts, I thought I’d ask if folks have any other suggestions, before I try going that route.

Probably not. There’s a big difference between what you consider “noise” and Audacity. Noise is not “whatever I don’t want.” The sound behind the singer is not noise. This interference is constantly changing and noise can’t do that. We can’t get rid of noise in a moving car for the same reason. It’s changing.

There can be no difference at all between the work you use as a profile and the interference you need to remove. Just in the several seconds I listened, the background noise changed significantly, sloshing back and forth like waves at the beach.

That and there’s a rule that the noise needs to be significantly lower than the show. If it’s not, you’re either going to get improper noise removal, noise stuck inside words and phrases, or vocal performance with martian, outer space voices. Sometimes two of those at once – and that’s if you do everything else right.

And it’s convenient if the original work is not internet compressed. That makes it worse.


Those are two valid tricks, by the way. Amplify the noise sample to -1 (not 0) and copy/paste so it’s longer and the profiler has more to chew on.


I was afraid that would be the answer, but it is what it is. Thanks!