there's an artistic effect I keep seeing that only works with generative AI. It seems to be alternating traditional transforms (translate, rotate, scale, perspective, color shift, take your pick of anything you want) with diffusion steps to bring the resulting image back to the same semantic prompt every frame. Looks trippy, like the image is moving, yet also staying the same (like a Shepard tone), and also constantly drifting in details because AI's not that good.