You can use DALL-E and other models to make pixel art ("as pixel art"), although it can both be overkill and hard to get consistent results that you'd put into animation. I'm guessing that starting from more of a video model and then converting to pixel art could be better. Although it's also non-trivial to turn "realistic" video into convincing animation.
I pay good money for a specialized machine learning algorithm that can take a pixel art character, and then generate all the animated sprites for it.
I actually tried to get Dalle to do this, And it made like three good sprites in the rest were just broken. But it was so strange, because you could see it was still organized as a sprite sheet, it's just the sprites were useless.
I think the practical applications of this technology will be hyper specialized models for specific purposes.
