I've seen AI image generation models described as being able to combine multiple subjects into a novel (or novel enough) output e.g. "pineapple" and "skateboarding" becomes an image of a skateboarding pineapple. It doesn't seem like a reach to assume it can do what GP suggests.