Can image generation do Jojo-style poses and composition well?