LoRA for Artemis/Tigress from Young Justice.
For Pony using @ 1 weight. Pony version doesn’t have tigress costume.
For v2.5 recommended weight between 0.8-1; can go higher or lower depending on additional LoRAs, etc.
basic outfit: green mask, green costume, sleeveless, midriff, arrow logo, green pants, knee pads
tigress outfit: stylized tiger mask, orange and black bodysuit, hip pouch, choker
school outfit: blue blazer, white shirt, necktie, skirt, preppy
(alt hair): shoulder-length hair, black hair, (or just put ponytail in negative)
(other): bow, quiver, sword, weapon
Generated Batch Size 6, and picked one seed. Then used Hi-Res Fix (Upscaler: 4x-UltraSharp) 2x —> Extras (Upscaler: 4x-UltraSharp) 2x
Description
Tigress costume can be prompted
overall better
Trained on AnyThingV5
FAQ
Comments (3)
I've been attempting something similar to this with a character who has multiple outfits. 2 'civilian' looks and 2 'superhero' looks. But elements from one keep appearing on another. They all share the same activation token and then another token to describe that particular 'look'. I don't suppose you could share at least 1 example of each costume from your dataset? This shows a lot more flexibility and less liklehood in having elements from one costume appear in another.
Example for 2 outfits. WD14 captioning got the general gist for clothes, but I added some details.
Yeah that was something I was wondering about as well before making this. There’s probably a better way, but my dataset for this was 65images. Maybe 1/4 tigress outfit, 1/4 green arrow costume, 1/4 preppy uniform, 1/4 nsfw. Couple others with like 2-3 images as well like a dress, jacket, turtleneck. But detailing all those means it should be compatible with most outfit models (like in the showcase images)
The mask I was kinda unsure on, but putting “stylized tiger mask” & “stylized mask” in the negative prompt when trying to generate the default outfit seems to work pretty well. Forgot to add a pic in the showcase.
@novelProphet Thanks very much! I'll do a rerun of my dataset and see if I get a different result. Mine are probably too large compared to your setup. I have 40 in one and near 70 in another.