Actually is a combination between, image, LoRA and prompt. I can give you some examples, try and tweek them yourself 🙂