This was produced in stable diffusion by putting all the frames into a 3x3 grid and processing them as a single image for consistency, using image2image and the softEdge control net. Then a pass through RIFE to double the framerate.
That's pretty darn freaken good actually! There's some inconsistencies, but the art style lends itself to a more jittery style of animation like this. This is where AI is most useful, filling in the labor-intensive blanks and enhancing an artist's hand-work.
