Steps

1. Record piano video
2. Use Neural Network to do body segmentation (video matting, https://peterl1n.github.io/RobustVideoMatting/#/) i.e. delete background
3. Run opencv (with python) to generate simple graphic engine
4. process audio to send signal to graphic engine
5. use IMovie to put everything together

Voila