Credit to the author Jin-jin.
He draws kind of hyper-muscular/shredded characters, mainly Felicia, with thick thighs and ridiculous proportions.
Common activation tags: 1girl, breasts, solo, muscular, long hair, thighs, muscular female, huge breasts, abs, thick thighs, large breasts, short hair, looking at viewer, sweat, cleavage, nipples, gloves, black hair, leotard, blue eyes, 1boy, dark skin, thighhigh, ass, nude
Description
FAQ
Comments (7)
Hm, from what I can see, you didn't include the generation data in any of the examples, or provide instructions about recommended strength or any potential keywords. I'd recommend updating the description and providing examples with data attached.
In a quick test run I did, strengths of 1 and 0.9 resulted in overbaked images, and 0.7-0.8 moved things into the usable range.
Maybe i need to start using activation words... Either way, i added common activation words to the description. Mind giving it another try? See what the words are, they are ordered from most common to less common, but muscle-related ones may trigger the lora more.
I posted the tribe prompt lineup now. Looks like it worked fine for the pics that had "muscular" in the prompt, and was done with lora strength 0.8
Hmm. perhaps it's something about your prompts that makes it not very close to the original artist style i think. I will definitely cook this lora again with a bigger dataset and an activation tag though.
@johhnypotato Well, I do use AutismMix instead of the base Pony Model. That might contribute a bit to some style drift. Your Lora still provides nice muscles and recognizable style elements, in any case.
Also, the prompt isn't individualized to work out the style of this artist - it's a standardized lineup of slender+muscular, male+female tribal villagers of a spectrum of species, in order to get a baseline of what each style lora does when confronted with it....
I've no doubt that by adjusting the keywords in the prompt, the style could be brought closer to the artist. This is another reason why visible prompts on your example pictures are important, as people can then have a look under the hood to see what you did to generate that style closely.
@Wahn129180 For some reason it doesn't import my prompts from Comfyui.. I added the common prompts i used, hopefully that helps. I'll be adding trigger words to my future LORAs.