Captioning a Vase LoRA with an LLM, but I'm doing it twice in order to get the "correct" output.