The idea here is just to support any Image Generation model. Stable Diffusion would be a good start.