AI tends to create extra body parts, can we limit that in the image generation to limit having to re-edit the scene.