The latest example is only a research paper for now, but a very impressive one, letting users simply drag elements of a picture to change their appearance. The interface is exactly the same as traditional image-warping, but rather than simply smudging and mushing existing pixels, the model generates the subject anew.
Sort: