Video models might learn unrealistic patterns for movement.

Video models are fed tons and tons of examples of movement.

But animation and video games don't do physically real movement, they do movement that feels right.

That creates a structural bias in the data… small but consistent, so even amongst noise it should pop out.