啊?现在关键词已经不太会发生混淆了么?是有新的插件还是有啥特殊的输入方式?

看你要求有多高，只是年龄差的话光辉系模型很容易做到，用age difference提示词加上相应描述就行了，如果是要求出的图都严格按照一定的年龄比例且前后一致的话，单靠提示词很难做到，只能自己炼金了，而且准备数据集也很麻烦。以下测试模型waiNSFWIllustrious_v110。
age difference, man and girl, 男人和女孩
提示词：highly detailed, detailed face and eyes, 1boy, short hair, necklace, print shirt, jeans, sneakers, backpack, 1girl, twintails, camisole, skirt, age difference, man and girl, side-by-side, outdoors, cityscape, full body, anime, masterpiece, best quality, amazing quality

age difference, onee-shota, 大姐姐和小弟弟，这个成功率很低
提示词：highly detailed, detailed face and eyes, 1boy, short hair, necklace, print shirt, jeans, sneakers, backpack, 1girl, twintails, camisole, skirt, age difference, onee-shota, side-by-side, outdoors, cityscape, full body, anime, masterpiece, best quality, amazing quality

danbooru.donmai.us/wiki_pages/age_difference上还有一些组合可自己尝试下，另外还有体型差身高差等等，不知道光辉模型是否训练了相关概念，没训练过的就画不出，只能靠LoRA了。

试试体型差这个词条

身高差

想稳还是推荐开提示词分区，给小孩头上单开一个空 分区，身高就压下来了