RIP onlyfans.
(twitter.com)
You're viewing a single comment thread. View all comments, or full comment thread.
Comments (54)
sorted by:
What I've learned is that the paper is only real when the code is released. Otherwise it's just smoke.
It's kind of cool how peer review is being pushed by the wayside by this. If the code is released and it works, we quote it.
Look at me. Look at me!
I'm the peer review now.
It's such a joy to watch academia squirm from that.
The 3D images seem jittery, and have clipping, the humans do not. That alone makes it suspect to me. Human model, actual physics, 3D models clearly guided by magnets.
I was more surprised about the cloth movement and how the body properly moved to that.
Which is why I'm comparing it to the 'drawings', which seems to actually be 3D models. Why would one fabric move realistically, and the other like a 3D model?
To steel man it:
Maybe it could have learned to animate from real photos with comparison to real videos with realistic clothing motion. And learned to animate cartoons from 3d animations with shitty clothing physics.
Well, here's hoping. I still love playing around with AI, I've enhanced my own images I drew with AI before, I'd like to put some motion to it, too.