There's definitely some manual curation. They even have assets that are clearly being placed into the scene, like the boats moving in the water.
You could theoretically generate all possible animated sprites (birds, people, vehicles, etc.) for use in any given cityscape, but automatically segmenting/detecting the different valid "paths/waypoints" where those assets should be placed would probably end up looking pretty janky.
Honestly I was blown away, it is one capability that blindsided me. I haven't been this surprised by a capability lift in a long time.
- parallax, which might be some clouds moving in the background.
- weather effects which is mostly just an overlay
- animated sprites that interact with, or at least move along, the different sections of the static scene.
It'd be neat to see how well it could recreate some of the larger set pieces like Westminster palace.
I've definitely seen some pretty impressive automated AI voxel stuff (and that was 6 months ago) and that's having to deal an extra dimension. [1]
Personally I still don’t see a situation in which you could do it completely hands-off though. I think you’d end up with some pretty good scenes and some really terrible scenes requiring post-adjustments. It would be an interesting experiment at any rate!