Artist Style LoRA for Lluisabadias

Description:

This is an artist style LoRA of Lluisabadias.

Training Settings:

• Optimizer: Prodigy / Cosine

• Base Model: Illustrious v1.1

• Final Model: NovaFurry v6b

• Trigger Word(s): 11uisabad1a5

• Training Steps/Epochs: 3063/33 (Clip skip 2)

• Dataset Size: 35

• Recommended Weight: 0.75-0.95 (0.80)

Sampler Settings:

• Steps: 30

• CFG: 4.5

• Sampler: Euler_a

Style positive design trigger words:

11uisabad1a5, masterpiece, best quality, amazing quality, aesthetic, cel shading, hard shading, lineart, crosshatching, outline

Character negative design trigger words:

Notes:

Seems to be slightly overtrained, even at 3063 steps. I may review a new set of training on this model and update with a new version, but 0.8-0.85 seems to work nicely. Some close-up images may produce better results at 1.0.

Reptile species seem to favor a more dragon-esque look, so you may need to be more specific.

Suggested Resources:

https://civitai.com/models/1001945 - (Detail Slider IL)

https://civitai.com/models/503815/nova-furry-xl - (Novafurry v6b)

Please remember to follow/support the artist if possible.