Baby Bonnie Hood (Vampire) 布丽塔 小红帽（恶魔战士）

tag:1girl,apron,capelet

She is a girl who loves to laugh, please use more tags like grin, evil smile

她是个爱笑的女孩，请多使用grin,evil smile之类的tag

————————————————————————————————————————

I would like to give you a suggestion. With the increase of lora models, many characters have multiple lora. Generally speaking, the file size of lora is determined by the dim parameter, such as 8M 18M 36M 72M 144M and so on. The larger the lora, the better the effect when switching to a large model, especially when an anime character is switched to a live model. According to my test, if you only make animation pictures, about 36M lora is enough, if you want to make real-life pictures, at least 72M lora should be used. Of course, the lora produced by some experts can appropriately relax this range.

给大家一个建议，随着lora模型变多，很多角色都有多个lora，通常来说，lora的文件大小由dim参数决定，比如8M 18M 36M 72M 144M等。体积越大的lora在切换大模型时效果越好，尤其是动漫人物切换到真人模型中使用。按照我的体感，如果你只制作动漫图片，36M左右的lora足够了，如果你想制作真人图片，至少要使用72M的lora，不然特征会丢失很多。当然，由一些大佬制作的lora可以适当放宽这个范围。比如上千张图片制作的10M大小的lora，动漫人物也会相当还原，但是通常情况下，如果你是跑真人图片，效果未必有3,40张图片制作的144M的Lora会更好。