Instead of just having one model as an option to generate images, why not have multiple to give the users more of a control of the output of the image generation?