SDXL Asari Lora

This LoRA has been trained on top of the SDXL 1.0 with different pattern. It was trained on a diverse dataset of 80 images of asari and 10 of human characters in the hopes that it would help flexibility.

The other Kohya settings can be found in the JSON metadata for the LoRA.

There was some nudity in the training set, but that was a minority.

Versions

Trigger words

I captioned the images, so you can control the skin color well.

Limitations

Notes

Previews

v0.1 (2030810-2)

Prompt examples

Unrealistic

Grid of samples Grid of samples Grid of samples

Realistic

Grid of samples