>>59833545I did the same thing you did, except I spent a lot of time trying to line up a sample of the BGM I grabbed off their stream to invert against the part where they talked. The picture here isn't lined up all the way though, but it was basically the same.
This was actually the final result of lining it up, from the one I posted
https://litter.catbox.moe/sgb8z3.wavYou can see it isn't clear. So again I just went through that .wav I made and literally silenced the audio where they aren't talking and trimmed it down to what I posted. It still isn't perfect, but it was my first time trying to do this.