If you have the capability to write new code off of research papers, we know how to produce arbitrary new video containing both real and novel humans, animals, plans etc, new behaviors and very large environments. https://dellaert.github.io/NeRF22/ more non-scene-generation stuff from that conference at https://blog.roboflow.com/cvpr-2022-best-papers-and-highlights/