I'm trying to do a favor for a young lady in Pakistan. I offered to clean up the recording and add some instruments. For the past few hours I've tried nearly every plugin in Sonar to get rid of the acoustic guitar in this file. I realize that a steel string acoustic and the inherent harmonics make it nearly impossible to isolate and crush the sound without also hurting the vocal. However, there are some pretty smart folks on this forum.
 
I was trying to get most of the guitar out, increasing whatever vocal was left, and then combining the two tracks in Sonar. The results have been rather dismal.
 
If you can do it, please email me the Sonar project file and I would be extremely grateful. (terrydelaney@outlook.com) I've been trying to get her to send me a pure vocal track, but like I say, she's in Pakistan. Not exactly the hub of the music industry.
 
(The file is in WMA format for ease of downloading. The original is in a standard 44.1 WAV format.)

If you have R-Mix it will handle it. 

I am on my cell now, but if they are panned the same (both centered), the frequency ranges can overlap enough to make this extremely difficult. R-Mix (from X2) may not cut it either (no pan or frequency separation).

I chimed in on this so I remember to check it out when I can. Do not get to hopeful of such things (is often better to bite the bullet and retrack them).

I just got to take a look at this and there is another issue at play. The guitar is louder than her voice, so there is frequency masking occurring which makes any separation nigh impossible.
 
Because there is a lead in on this, I tried using Audition to preform noise reduction on each chord one-by-one at a gentle reduction. This preserved the voice until I got about 5 chords into it, and left a squeal artifact (also due to how much this was digging into the voice frequencies). One caveat to any "noise reduction" is that the signal must be stronger than the "noise" (unless some other separation criteria is available).
 
In all honesty, if she has any means to retrack this (can sing while listening through headphones), that will give a better starting point than trying to separate out this track. Even with consistent environmental noise, the vocal can be separated with sufficient signal-to-noise ratio.

Have her record over the existing track at a soft whisper, with the shower running, the words bomb, jihad, terrorist cell. When the NSA finds this track inbound from Pakistan, they will use the kind of technology you need to pull out the vocal. Getting the resulting clean track from them will be the next problem.

What you might try is to bounce a mix exaggerating the offending instrument via EQ.  Let Audition sample that track for noise reduction and then in the restoration function of Audition apply it to the original track.  Worth a try.

Yeah, looks like she's gonna have to do what mettalus is suggesting. She usually visits her alma mater here in Boise, Idaho once or twice a year, so I'll grab her and get her into my studio.
 
@slartabartfast: I read an article somewhere that claimed the NSA can isolate a single conversation from a noisy crowd. Seems impossible to me, but I'm just a hack keyboardist.

Woo-HOO! She dropped two clean audio files into a Google drive (separate voice AND guitar). So now I can get to work.

Hehe, the REAL solution