Update: I have remade the model, and now it is a LoHa model, so there are some differences in how it is used compared to the previous LoRa version.

Note:

This is a LoHa model, and you need to install a1111-sd-webui-locon extension to use this model.

And don't use sd-webui-additional-networks to load this model.

Place the model in the path '.\stable-diffusion-webui\models\Lora' to use it.

Compared to the previous version, it is now easier to change clothing and generate NSFW images.

Suggest weight:around 0.9

Suggest prompt:

dionags, dionaoutfit,

navel, shorts,shirt,belt, hat,bare shoulders,detached sleeves, gloves,pouch

Also, many tags are optional and can be removed as needed.

If you want to change to other clothing, you only need to keep a basic dionags tag, which contains the basic characteristics of the character.

However, there are also some small issues, but they are all very easy to solve. For example, sometimes the background may become white. You can simply address this by writing white background or simple background in the negative prompt.If the tail is missing, add it to the prompt.

(However, it has to be said that the new version seems to have inferior performance in some details of generating default clothing compared to the old version. But its generalization performance has improved significantly. I have been adjusting the hyperparameters of the LoHa model during this time, but I have not found the best ones yet.)

The following is a description of the old version:

This is a description about the old version and is not applicable to the new version.

Of course, I strongly advise against downloading the old version.

trigger word:

diona\(genshinimpact\)

, hat,gloves, detached sleeves,shirt,belt, shorts, navel

hat and gloves are optional,remove these tag to take off hat and gloves.

The model‘s generalization is not so good.After testing,it's hard to change clothes.But change expression and pose is feasible.if you really want to change clothes,reduce weights to 0.2, adjust prompt and add negative prompt belt,bag.it can make changing clothes easier.

I suggest around 0.5-0.8 weight (test on anything v4.5),if you want generate NSFW image ,Please reduce weights to 0.3.

trained on novel-anime-full model.