--Updated-----------------------------------------------
当时训练模型还有很多不成熟的地方,现在测试起来问题还挺多的
比如概念的融合,虽然有各种方法来减轻这种问题,但这是无法完全避免的问题。分开训练可能才是最好的方法。
这个模型无论是泛化度还是拟合度都应该更好,相比最初的模型,它唯一的缺点是需要更多的触发词唤醒。
At the time, there were still many immaturities in the training of the model. Now when testing, there are still quite a few problems.
For example, concept fusion - although there are various methods to mitigate this issue, it is an unavoidable problem. Training separately may be the best approach after all.This model should have better generalization and fitting capabilities compared to the initial model. Its only disadvantage compared to the original is that it requires more trigger words to activate.
Strongly recommend enabling Hires.fix, the Lora weight can be set to 0.5-0.9. If there are requirements for changing clothes, it is recommended to set the weight to 0.6 or lower.
Default skin:arknights_mint,long hair, black thighhighs,white dress,grey hair,blue eyes, black ribbon, cat girl, hair ribbon, ribbon, bangs,ahoge
"yuexiao"skin: arknights_mint_yuexiao,japanese clothes,hair flower,wagashi,sash, long sleeves,obi
强烈建议打开高清修复,lora权重可以设置为0.5-0.9,如果有更换衣服的要求,建议降权重设置为0.6或更低。
原始皮肤 arknights_mint,long hair, black thighhighs,white dress,grey hair,blue eyes, black ribbon, cat girl, hair ribbon, ribbon, bangs,ahoge
月宵皮肤 arknights_mint_yuexiao,japanese clothes,hair flower,wagashi,sash, long sleeves,obi
Description
FAQ
Comments (7)
居然連新皮也做了!佬!
第一张大图好可爱啊,想抱起来狠狠的rua。能问问老哥你出这张图的时候用的什么model吗,我爱死这个闪亮的眼睛了
Anything3.0应该会有最相似的效果
薄绿的前发果然还是很难把特征跑出来,素材比较少,我之前也跑过薄绿的lora,但前发就是要散开(其实就画师自己的几张图刘海特征也不完全相同....),感觉要跑出前发的话可能得用几张立绘跑出过拟合再用AI继续出素材跑了吧....
i will never publish another lora👍
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.
