I guess the issue he's referring too is what you see in LA Noire, great facial animations but a disconnect with the rest of the body. By improving one part you highlight the flaws of another part.
You get the body moving right, then you start noticing that the clothes or the hair doesn't work exactly right, or that sweat is not behaving naturally.
Interactions between 2 characters is even more problematic. They can't even get it right yet on film when actors have to act against a prop. You can always spot that the actor is not actually looking at the thing you see on screen.
But you can still move out of the uncanny valley, it doesn't have to be perfect for that. Things might not seem quite right, but that doesn't mean the characters freak you out.
But what does he mean by this?
"I think Beyond is trying something interesting. For current gen, it is going way further in terms of quality, but it is too complex to implement on next-gen."
Why is it too complex for next-gen? Too complex for their real-time engine or does he mean something else?
What I would like to see next-gen is characters that can walk through a doorway and past/around another person normally, it's about time for that.







