During covid I needed to setup remote piano lessons. I tried combining two videos to show hands on the keys as well as the face but at the time merging the two videos added too much latency.
I solved this by only using the top down camera and then adding clear perspex (from a plastic picture frame) at 45 degrees in front of the raised laptop and also pointed lamps at the piano players face (the effect is called Pepper's Ghost). This meant the teacher could see the hands on the keyboard and the players face with one camera. It also meant the piano player would appear to look directly into the camera - like an auto cue.
I'm posting here in case it helps someone else.