Yes, it's asking SD for an image with some set of characteristics, and SD has some notion of what is plausible from what it saw during training.