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.

Suggest weight: around 1.

Trained on 1,777 ukiyo-e prints from artists such as Kitagawa Utamaro, Katsushika Hokusai, and Utagawa Hiroshige.

During my training, I did not process the training dateset, so the results may not be optimal. If the results do not meet your expectations, please also forgive me.Therefore, I have uploaded the training dataset. If you are interested or think you can train it better, you can download it and train it yourself.

Regarding the v1 and v2 versions, I used a too high learning rate for v1, which caused some overfitting and resulted in not so good results for generating portraits. The v2 version should perform better overall, but in some cases, v1 may actually work better for generating certain scenes.

Actually, you can also download the v1 version because it has more vivid colors and resembles the ukiyo-e style

Based on novel-anime-full model.

Tested on anything-v4.5.