You'd have to be r-slurred to write a story like that

its a combination of using one type of model create a story board of generated images, then you finish it up by running it through another model that "turns it into a gif" to animate it starting from the input / seed image.

It requires an insane amount of VRAM so all consumer cards are limited to generating max about 5 seconds of decent resolution animations. On the bright side, most media is already ADHD super-jump cuts anyway on average about 2 second clips stringed together so it ends up working for this sort of gay shit.