AI虽然能识别一些动漫人物的词条，不过都是一些非常有名的角色，所以对于大家喜欢的cp一般只能使用各种词条描绘来尽可能的还原，可能需要跑很多张，但是中间出的副产物也别有一番风味









前面两张是神雨图，3，4两张是跑的梦结和riri，剩下的都是跑图中间出的副产物（最后一张本来是想跑安达与岛村，不过岛村的头发颜色描述不好就没多跑了）
咒语
(best quality), ((masterpiece)), (highres), original, extremely detailed 8K wallpaper, (an extremely delicate and beautiful), Blush, open mouth, smile , 2girls, kiss, Church, Yuri, wedding, garden, ((detailed face)),beautiful detail eyes, hair fluttering in the wind, messy hair, lens flare, soft shadows, real lighting, detailed lighting, garter straps and garter belt,(deep blue hair and flat breasts and very long hair and blue eyes and detailed wedding dress and young girl ),(short pink hair and pink short hair and side ponytail and four-leaf clover hairband and four-leaf clover hair ornament and four-leaf clover scrunchie and flat breasts and pink eyes and detailed wedding dress and Little girl ),(floating glass fragments), ((colorful refraction)) ,own hands together, on side ,eye contact, caustic, face to face
这是跑梦结yuyu图的时候的tag，跑别人就更改一下描述就行了，负面tag用通用的就行了，但是最好在开头加上一个looking at viewer，这样出的看向你的图就会变少