IT'S CAFO TIME. Four hands, two tracks. ALL AUDIO is live (NOT sped up!) and unedited from the left video. It's the playback from the right video, with what I'm playing on top.
As for video:
First, I recorded the video on the right, in one take, but in two separate tracks (my piano doesn't have enough memory to record the whole thing). The first track ends after the 2nd section (the clean guitar picking in 13). At around 2:30, there are about 2 seconds of time in between tracks, where I set up the 2nd track to record, that I edit out in order to keep the two videos synced. That is the only part of this you're not seeing in the video.
Second, I recorded the main track, all in one go, switching from one backing track to the other without any editing (you can see a highly rehearsed button-pressing routine going down around the 2:25 mark).
Split-screened the resulting two videos and aligned everything with the left audio track.
A little convoluted, but after a lot of experimentation I felt this was the best way to keep the "one performance" feel intact and do the final audio in one take.