A related paper by the same lead author:

"MotionCLIP: Exposing Human Motion Generation to CLIP Space"

https://arxiv.org/abs/2203.08063

Would anyone be able to explain how the two techniques are related?