What is “predicts what should be under your brush” supposed to mean? I don’t see it it anywhere in the paper or the video.
I think a clear way of saying this would be “the algorithm tries to realistically modify the overall image to accommodate the user’s changes.”
That is, if the user paints black over a face’s forehead, the algorithm finds the closest real image that can match those changes, such as black hair in bangs over the forehead, and modifies the image to incorporate that hair.
