I’m hoping someone can describe a technique for doing this.
I need to improve sound separation on a stereo recording.
I don’t have the original instrument tracks to filter/eq.
This is the file I’m working on, in Audacity.
What’s the best way to manipulate the frequencies in this situation?
Separate what from what ?
Completely Isolating instruments from final mix, (when you don’t have the original recordings of the individual instruments), is next to impossible.
If the instrument you want to cut or boost only occupies a narrow band of the audio spectrum it is possible to do that using the equalizer, but the sounds from other instruments or singers which fall in that narrow band will also be modified.
Try searching this forum for the word “isolate”, e.g. … https://forum.audacityteam.org/t/how-to-remove-music-and-keep-vocals/11524/1
The first thing is to upgrade to Audacity 1.3.13 as this new version will give you a lot more control (much better tools to work with).
You can get Audacity 1.3.13 from here: Audacity ® | Downloads
The “Plot Spectrum” window will generally give a more useful display if you set the “Axis” to “Log frequency”.
Could you try to describe in a bit more detail exactly what you are trying to achieve.
Hi Trebor. The overlapping frequencies of a rhythm guitar and lead guitar. There’s only two instruments in the recording. I can’t record the part again, unfortunately.
Yes, I’m aware of the limitations, using a stereo mixed file.
I’m after a technique/method to identify, and begin reducing, the conflicting frequencies as best I can.
Steve, I’m using Audacity 1.3 beta. The mix is a loud rhythm guitar using heavy distortion and a lead guitar using sustain. I’m trying to reduce the amount of noise where the frequencies overlap. I think I need to identify the side band frequencies of each instrument, but I’m not sure how/where to do that just looking at the analysis spectrum.
Oh yes, I see that. Is that on Windows XP? (not that it matters)
You can increase the detail by:
a) Making the window bigger
b) Increase the “Size” parameter
c) Set the “Axis” to “Log frequency”
Unfortunately I don’t think that it will help much as the two guitars are likely to be mostly occupying the same frequency range, so if you boost or cut one, then it will do the same to the other. The “distortion” sounds will be mostly in the higher frequencies so you may be able to make a subjective difference by reducing the level of the higher frequencies, but this will be at the expense of making the entire recording sound a bit dull.
You may be able to achieve a bit more separation between the instruments by widening the stereo field. There’s a plug-in called “channel mixer” that can do this, but I don’t expect it to make a radical improvement. Missing features - Audacity Support
Win7. I only installed Audacity 1.3 today. I still have 1.2 installed.
Does it look like I’m getting data settings for 1.2 in 1.3? I’m wondering why you asked.
Ok, thanks for that. And also, I installed a vst plugin called voxengo span (ocsilliscope/spectrum analyser). I have it’s .dll in the vst plugins folder, but audacity’s search doesn’t load it?.
I wondered if I isolated them, one at a time (destroying the other), I’d end up with two tracks to remix. But without removing the overlap, I’m probably not achieving anything.
Yeah, I imagine so, without decent separation in the first place. But thanks for the heads up.
If you do two frequency analyses: one of the lead guitar on its own, one of the rhythm guitar on its own, then compare the graphs.
Unfortunately I suspect they will both have very similar frequency content , if so it won’t be possible to emphasize one using equalization.
[If it was a bass and a lead guitar you wanted to separate that could be done with equalization]
http://manual.audacityteam.org/man/Audacity_Selection
This is the problem, because of the amount of distortion overlap and that it’s already mixed down. If I could analyse each one separately, I would just remix.
Yeah, I’ll have to go with what I’ve got for the time being. 
So the guitars are both playing continuously throughout the entire track ?, no point where the lead stops and there is only rhythm ?
If there was a bit with only rhythm you could select and analyse it,
maybe even use the bit with rhythm only as the “noise profile” for “noise reduction” in an attempt* to make the lead more prominent.
[* don’t get your hopes up]
Ok, I hadn’t thought of that. Worth giving it a try. Thanks Trebor, much appreciated.
If it is a stereo recording, the technique described in http://www.endolith.com/wordpress/2009/11/22/a-simple-fastica-example/ can be tried. If you achieve good results, please tell us.
In order to successfully separate X sounds fastica needs exactly X channels. If you have stereo you can only separate two sounds. If you have one more sound in the field it won’t work at all. My track has three guitars mixed across the stereo field.