[a.i.]It's never going to be as good as the human input
That seems like more of a philosophical question.
Feeding AI a random disaster planning scenario (most governments give them out to anybody that wants to read them), picking a random disease from the book of exciting diseases for your medical show, picking a what if scenario to stretch out from the handful of sci fi or fantasy tropes... the standard thing the majority of these writers do (and probably fail at doing well).
Picking one of the plot types
https://www.masterclass.com/articles/a-guide-to-story-archetypes (for those unaware you have 7 basic story types that basically everything is a version of).
Picking a writing style or setting accordingly...
That will get you a serviceable thing already, and some of the models are getting token counts and memories of previous actions intp the quite ridiculous camps (not sure where the hardware limits might wind up but still likely a while you yet), considerably better than if you only tried chatgpt when it went viral a few months back. Might want someone playing editor (curator might also be a term) but that is often an awful lot easier.
Voice work is there already (got very nice emotions, songs, languages... and fitting it to someone blathering in their webcam mic is even better).
Sound effects are there already but not quite as good as voice.
Music is doing very well and almost there for most purposes.
2d stills are better and better. Background art is already actively used.
Moving replacements of 2d stills is a lot of work but nerd in a bedroom level. More than good enough for a cartoon.
3d models is a bit out but not much. Should not be too hard to couple with mocap either (or indeed fake it if you can't be bothered to get a suit or paint dots on your face).
@
realtimesave:
