P3的反推
1girl, bow, flower, dress, long hair, bonnet, long sleeves, twintails, green bow, blonde hair, smile, white background, full body, green footwear, frills, open mouth, looking at viewer, shoes, solo, very long hair, socks, bangs, puffy long sleeves, green bowtie, rose, :d, lolita fashion, simple background, stuffed toy, standing, bowtie, grass, puffy sleeves, stuffed animal, yellow eyes, white flower, orange eyes, frilled dress, hat, blush, green flower, capelet, hair bow, green dress, hair ornament, white socks, floral print, striped bow
测试下来 可能是grass

一个比较省检索力的方式，比较好的找关键词以及参数还有lora配置的方法是，去c站随便找几个大模型或者lora的下面生成图片（必要时点开图片的信息页，会显示用了什么大模型或者其他lora，某些情况下会不显示lora或者大模型，甚至有的图片没有符合sd生成图片的关键词，这点要记住），你试着找到好看的图片，以此为基础丢入png信息那个模块里，转到图生图

真好看，蹲个参数