Well put. Honestly the actor part is mostly solved by now, the tricky part is depicting any kind of believable, persistent space across different shots. Based off of amateur outputs from places like https://www.reddit.com/r/aivideo/, at least!

This release is clearly capable of generating mind-blowingly realistic short clips, but I don't see any evidence that longer, multi-shot videos can be automated yet. With a professional's time and existing editing techniques, however...