Is working with an agent different than constantly editing an image with fine tuning prompts? Both can be entirely generated.