Follow me on X/Twitter for exclusive content: https://x.com/Citron_Legacy 😁
Please checkout all my Style Loras: https://civarchive.com/collections/10766871
Please check out my other Western Animation Loras: https://civarchive.com/user/CitronLegacy/models?tag=western%20animation
I think my Teen Titans style works better than this one.
Asami Sato - aaAsami, light green eyes, red collar, black jacket, long hair, black hair, green eyes, jacket, belt, makeup, lipstick, black dress,
Korra - aaKorra, brown hair, ponytail, hair tubes, blue shirt, sleeveless, bare shoulders, armband,
Green Opal - aaOpal, green eyes, short hair, black hair, green dress,
Red Opal - bbOpal, green eyes, short hair, black hair, red military uniform,
Lin Beifong - aaLinBeiFong, armor, short hair, green eyes, grey hair, black uniform, makeup, scar, crossed arms, lipstick,
Suyin Beifong - aaSuyinBeiFong, short hair, grey hair, green dress, neck plating, lipstick, red lips,
Description
FAQ
Comments (3)
Thank you for this, cool to see you can train so many characters at once!
Any chance you're planning to make one for the original series?
Glad you like it!
Train multi-character loras on Illustrious is nice because it doesn't overcook and also has way more understand of characters/clothes/concepts than any other base model.
Sometimes its as easy as throwing a bunch of different character datasets into one big dataset.
Thats a nice idea: I like the original much more than Legend of Korra.
Someone gave me the datasets for Korra characters so it'd probably take me a long time to get all the datasets for each female character in the original. Katara, Toph, Suki, Ty Lee, Azula, and Mai right? I think there is a water tribe girl that turned into the moon but she probably doesn't have many images.
@CitronLegacy I didn't realise Illustrious was easier to train too, or that you could do multi char that easy. Is that also the case for male/female characters - like if you added Aang/Sokka/Zuko into the same lora?
In terms of characters, June would also be cool too if possible! A few questions about dataset:
- What res do you need to train at/do you only train on show screencaps?
- How many images do you need per character?
Maybe I can help with some of the dataset depending on what's needed