It took 2 days to train 4W steps of fusion in Kontext

If you are worried about blindly downloading renderings, please contact me

The generalization is stronger, the output is more stable, and the effect of changing clothes is more explosive.

After 5 iterations, the identity consistency still reaches 95%

And unlimited NSFW

And unlimited NSFW

And unlimited NSFW

Important things should be said three times

The workflow can be done with the official one

You can add 8-4 steps of LORA to the output

Prompt word: Take off the hee hee clothes (hee hee represents roles such as - girl, woman, man, boy, woman, etc.)

If you need to enhance the effect, please ask me for the corresponding LORA.

花了2天时间，训练了4W步融合在Kontext

担心盲目下载效果图找我

泛化性更强，出图更稳定，换衣服效果更炸裂

迭代5次后身份一致性仍达95%

而且无限NSFW

而且无限NSFW

而且无限NSFW

重要的事情说三遍

工作流用官方的就可以

出图可以加8-4步LORA

提示词：Take off the 嘻嘻 clothes（嘻嘻代表角色如--女孩，女人，男人，男孩，妇女等）

如需要加强效果，找我拿对应的LORA