(the technical difference is that “image generation” usually starts with a random noise pattern as the base image, whereas this started with an upload of the original. could be different models but idk if there’s an ethically derived model out there sufficient for this task)
