This LoRA is designed to generate casual illustrations that mimic the distinctive portrait style of "Elin" by Lafrontier, blending it with a soft, crayon-like aesthetic.
The training dataset is based on the "Elin Skin/Portrait Assets" portraits I previously generated. The creation of these assets wouldn't have been possible without the following incredible models:
Elin Portrait Style LORA
Illustrious Enhancer LORA
I would like to take this opportunity to express my deepest gratitude to all the creators of these models.
Description
Initial. (Experimental)
A LoRA weight of 1.5 or higher is recommended.
MOSS Image recommends detailed natural-language or JSON-based captions. Therefore, training on this Danbooru tag-based dataset is experimental.
LORA, MOSS Image Base, dim32/alpha32, AdamW8bit, LR 1e-4, cosine 2000steps, 124images, 1024² res.
Captioning: Booru tags only(including some broken English).