This LoRA is designed to create an anime watercolor style. It was trained on the illustriousXL model and tested on the waiNSFWIllustrious model. The recommended parameters are generally the same as the illustriousXL model.

Recommended parameters:

• CFG Scale: 3-4

• Steps: 30

• Sampler: Euler a

• Resolution: 832x1216 or 1024x1024

The training dataset for this LoRA consists entirely of AI-generated images, which has produced surprisingly good results. For certain prompts, there may be significant overfitting, which can be suppressed with negative prompts.