It turns out that with the proper tweaks you can tell Stable Diffusion something like "Take my avatar image, keep just the face and the hair, then fill in the background as though it's a movie-poster"
Then you can leave it doing hundreds of them overnight.
Then you can stitch like 1,600 images together in a video running 3 of them per second.
If you do all that, you get this:
https://dalliance.network/w/7s4K2pMoyWsXBxJu3r86wG
It's not even hard to do (now that the hard research is done by others).
On a linux/Mac console with Python3.10, a GPU, and a file called MyAvatar.jpg all you need do is:
> pip install imaginairy
> imagine --init-image MyAvatar.jpg --mask-prompt "(face OR hair){-9}" --mask-mode keep --init-image-strength .0000 --fix-faces --repeats 100000 "man wearing a space suite" "cover of life magazine" "film noir movie poster" "scifi book cover" "robots and spaceships" "super hero costume" "actor head shot" "gangster movie poster" "transformers movie poster" "professional uniform" "romance movie poster" "fantasy book cover" "ghost movie poster" "vampire book cover" "policeman in uniform" "firefighter in uniform" "movie poster" "a man in a space suit" "space suit" "space uniform" "star trek" "star wars movie poster" "batman and joker"
Then let it run generating images overnight before you stitch 'em together:
> apt install ffmpeg
> ffmpeg -framerate 3 -y -f image2 -pattern_type glob -i 'outputs/generated/0*.jpg' output.mp4
Video Embed:
[If flashing images are bad for you this is probably bad like that]