You can try my tool in
to isolate the center.
In order to do it properly, you have to rearrange the two files. Split them into mono.
Then, make a stereo file out of the left channels and also the right channels.
The center should now hold the common audio, namely the instruments.
You can then try to isolate the center in those tracks which gives two dual-mono tracks.
Split those agin and delete one channel from each. Make the remaining two mono tracks into the final stereo track.
There will probably still be a lot of vocal residual, especially where the lines share similar content.
You can use your first version (where the voices were clear) as a noise profile and apply it to the just made stereo track.
I wouldn’t expect too much though.