I've found that there is a dissatisfying lack of iteration, and modeling in a 3d sense of image generation. Like if Dall-e Spits out an image, I want to be able to move the camera location/angle. Delete certain features (like a freckle or extra hand), place/remove objects from the scenery, change the time of day etc... Both because I'd like to be able to create permutation sets ({Red, blue, green, #c71010} colored {dragon, hydra, medusa, zombie, werewolf}) but also because the specificity of what I'm imagining is hard to get right in a single prompt and I find myself setting up words till I get close, and then just generating 25+ times and picking the best output.
Kind of like someone else posted on this Reddit: https://www.reddit.com/r/artificial/comments/11ovh3j/iterati...