The LLM will. But the image generation model that is trained on a bunch of pre-specified tags will almost immediately spit out unrecognizable results.