Latest Update!:
Anima version is released. This checkpoint is really awesome. Since it understands natural language, I added some unordinary stuff like "girl is horny", "girl is naughty" etc.
OLD (Z-image):
Attempted to make a Z-Image-Turbo version.
THIS IS MY FIRST LORA FOR Z-IMAGE. Quality is questionable.
Either I'm doing something wrong or this model is just bad for nsfw or anime in general.
Not too bad though.
Update! (Pony):
New version is released!
Changes in this version:
Some technical changes
Improved outfits
Much bigger dataset (almost three times bigger)
Better pussy
She has no longer brown hair. It was a logical mistake in V1, it looked like brown due to the artstyle and shadow. Now, it's blonde hair. And shadowed hair depending on the situation.
Another Update! (Illustrious):
I made the illustrious version. Someone requested. Trained on NoobAI-XL.
It's kinda better than Pony. Not surprising tho. It's a superior model (for now).
Description
FAQ
Comments (9)
Can you take multiple hentais from this studio and make their ultimate lora style ? This would be awesome
Yes i can. I'll consider it. But expect something big like 400mb because... Technical stuff.
@Ringakiseki thats ok size, its like 2 standart loras
Look forward to this studio's style lora
Amazing Lora, I have a quick question tho. How do you make a Lora on the character and also the style? Like you've managed to get the Lora to look exactly like it came from the anime.
Glad you liked it, thank you. Regarding your question, it is very easy. Base model already knows there are different styles. So, basically adding ["name" artstyle] into the tags makes ai to recognize the style. Therefore, there are more things to learn for it. Tags are VERY VERY IMPORTANT. I'm gonna give you an example for this: Let's say we want to make a style for XXXX anime, whatever the name it is, then we add XXXX artstyle into the tags. The tag files, .txt files for example. Make sure you named your character as well. And tagged.
@Ringakiseki Thank you very much for answering, I really appreciate it. So let's say I want to make a Makima Lora but also want to keep the Chainsaw Man anime style, all I need to do is add a tag to every image in the dataset called "chainsaw man art style" right?
@ShinobiSpartan Yes. I recommend you to type as "artstyle" not as "art style". And tag Makima as well.
@Ringakiseki ok, thank you a lot again, appreciate it