[nopinsight]

Have you seen Veo 2 just launched by Google? Its quality and physics understanding appear far ahead of the competition.

https://deepmind.google/technologies/veo/veo-2/

Also, planning might be around the corner with test-time compute applied to video generation.

[wcarss]

There's also Genie 2:

https://deepmind.google/discover/blog/genie-2-a-large-scale-...

this one's entirely about world understanding with physical concepts etc. and less about photorealism, but it's really not hard to imagine a pipeline combining these