The DALL-E API doesn't give you access to the latents or model, so editing/consistency is much more difficult than it has to be. If you wanted to work within t

The DALL-E API doesn't give you access to the latents or model, so editing/consistency is much more difficult than it has to be. If you wanted to work within the current feature set, I think inpainting or 'uncropping' is probably the way to go. For example, generate a character until you get one that you like; generate a bunch of variants of that one face in various positions/angles/emotions (either through the variation feature or by reprompting); now, cut the faces or figures out of each one, and use them for each illustration as inpainting or uncrop with a text prompt describing that scene. So you'd generate your target girl, get a bunch of her faces (uncertain, sad, facing left, facing right), cut them out for an outline, and then inpaint/uncrop: "Shawna felt sad to be away from home and Mom all day"+[sad-face.jpg]+"a brown-haired young white girl in a gingham dress standing alone in a colorful cozy kindergarten classroom"; and so on for each scene.