This is a style LoRA compatible with both ILXL and Noob model series.

It was trained using reference images from 3D cel-shaded games and MMD visuals.

I use ComfyUI with node connections to call the LoRA, so there are no specific trigger words.

If you're using a different tool, you can call it using the <lora:~> syntax.

The base model used for training is ILXL, and the recommended LoRA weight is between 0.6 to 1.0.

For the example images, I used NAI-XL as the checkpoint.

Using a strength of 0.6 results in a more 2D-like appearance, while 1.0 brings it closer to 3D realism.

Although it was trained with ILXL 2.0, it seems that when using it with WAI-based models, the default WAI style tends to overpower the LoRA style, making it less effective...

The images below were generated using the same prompt and seed, with the only difference being whether the LoRA was applied or not.

Recommended tags are as follows:

koikatsu_(medium), game cg, 3d, mikumikudance_(medium)

If you want for certain styles, the following is also recommended (for a more 2D-like feel):

no nose , shiny skin