LoRA for Depresso, a dark element Pal from Palworld.
Recommended weight: 1.0
Main trigger word: depresso
For "default" look: blue body, purple hair, pink eyes, black tail
Feel free to adjust colors to get your own unique Depresso:
I swear, this is the last one.
Did minor tweaks to the dataset to fix issues that popped up in version 2.
Also changed network_dim to 16 from 32 to lower LoRA size.
Overall I'm very satisfied with results and I can say that it's a definitive improvement over the version 1.
Decided to redo this LoRA by reworking captions and tweaking dataset a bit. Not a major improvement (if there is an improvement), but seems to be more flexible, at least in some ways.
Known issues:
messy eyes
"collar" sometimes doesn't appear fully black
bizarre forest
Initial release
My first serious attempt to use LoRA training on this website. The model turned out to be surprisingly decent, despite the fact that training data was kinda a mess.
10 epochs, 480 steps;
Recommended weight 1.0;
Trigger word: Depresso;
blue body, purple hair, pink eyes, black tail for "default" look, colors can be changed (usually works well).
Minor tweaks in dataset in attempt to fix issues that popped up in v2.0
Also changed dim to 16, as default 32 seemed to be too extreme for a character like this.
1. 재게시된 모델의 권리는 원 제작자에게 있습니다.
2. 모델 원작자가 모델을 인증받으려면 공식 채널을 통해 SeaArt.AI 직원에게 문의하세요. 저희는 모든 창작자의 권리를 보호하기 위해 노력합니다. 인증하러 이동
