This is a style and (some) characters lora trained on screenshots from Justice League/Unlimited and other DCAU shows with the streamlined animated series designs.
Trigger word: dcaustyle
Strength: 1.0 (or lower if style is too exaggerated)
For the Pony XL v.6 version, it's most consistent when used with 2D anime or cartoon-based merges. I recommend Prefect Pony XL v2 which can get close to the authentic style.
For the Illustrious XL version, try using models like WAI-NSFW-illustrious-SDXL or Hoseki - LustrousMix. See the "Version 4.i" section below.
Use booru tags. Natural language prompting can work too.
Notes
Dataset was made mostly from screenshots from fancaps.net of JL/U, upscaled TNBA (after the redesign) and the movies that used the Timmverse art style. I put more emphasis on major female characters, but I included some men, random characters, and scenery. I avoided crowd shots or pictures where characters were overlapping, since that seemed to confuse the AI based on a previous attempt.
Good for solo images. May not be very coherent with complex multi-character images. As always, use detailer or inpainting to fix faces that are small in the frame. This was trained on civitai because I don't have enough vram.
Many of the frames from the cartoon are simple and at times "badly drawn". Maybe it was a bad idea to train this on "bad" drawings, but I wanted it to look authentic to the TV show. (Others have made loras that look more like Bruce Timm's personal artwork, such as: Bruce Timm - Portrait Sketch Style and Retro Cartoons, the Animated Series.)
Version 4.i for Illustrious
This version was trained for Illustrious on a similar dataset to version 4. Follow prompting instructions for Illustrious, using booru tags and "dcaustyle" and optionally the character prompts listed below. I got good results using it with the model WAI-NSFW-illustrious-SDXL or Hoseki - LustrousMix. It might not play well with NoobAI models. If you find models it works best with, please leave a comment.
Version 4 Pony
Trained on 330 images. New version with a slightly tweaked dataset and captions, and trained a little longer. I think this has better style and character accuracy than previous versions. Eyes and hands might still be questionable, so use inpainting or detailer to fix.
Version 3 Pony
Trained on 314 images and trained longer over all. I added more diversity of images split between: heavily-trained characters, less-trained/random characters, and a small amount of scenery.
I tried strategically removing some tags from some images. For example, I dropped basic hair and eye description tags from most images of characters that have more than 3 images. This is an effort to reduce trained characters bleeding into original characters. Will this work? I don't know. Some characters have been learned better, but some might be worse than v2. Overall though the style is stronger.
Some clothing tags for known characters were changed. See the Character section below for the current tags.
Version 2 Pony
Trained on 212 images, up from 128. Added more pictures of the featured characters I'm trying to train and random characters/costumes for dataset diversity. This version has more varied male characters too, though major male characters were not represented by more than a couple of pictures and will probably not be accurate. Sorry.
Some tags used for major characters were changed. See the updated character section below. Training was slightly longer with tweaked settings. (2 repeats x 17 epochs = 34 looks at each image.)
SD 1.5 version
I made this by request, just with mostly the same dataset as v2. It doesn't work that well. Use it with a 1.5 anime model. You'll need to use hires fix, inpainting, detailer, etc. to improve the images from this one. Tags should be the same as v2, except of course don't use Pony's tags.
Characters
This is meant to be a style lora, but I also tried to make it like a multi-character lora.
At this point, others have made Illustrious and Pony character loras for pretty much every JLU/DCAU character, so I recommend using those together with this lora for best accuracy.
Some characters were trained with about 9-15 images each, and can be generated with moderate accuracy. Try using the name and some of the booru tags that are listed below. (You don't have to use all the tags, just what you need to emphasize. Try also prompting "constricted pupils" if you want pupils, "no pupils" if you don't, or even "no sclera" as some DCAU characters like Superman don't have sclera.)
Batgirl: batgirl, light blue eyes, mask, red hair, black lips, two-tone cape, belt, black bodysuit, yellow gloves
Barbara Gordon: barbara gordon, light blue eyes, no pupils, red hair, black lips, black shirt, pleated skirt
Black Canary: black canary, blue eyes, blonde hair, sidelocks, medium breasts, cropped jacket, black leotard, pantyhose, cleavage
Galatea: galatea, blonde hair, bob cut, swept bangs, aqua eyes, gold belt, turtleneck leotard, long sleeves, cleavage cutout, gloves
Hawkgirl (Justice League season 1-2 masked outfit; you have to prompt "feathered wings" or it will leave them out): hawkgirl, feathered wings, orange hair, strapless shirt, belt, tights, red footwear
Hawkgirl (JLU outfit with no mask): shayera hol, feathered wings, green eyes, constricted pupils, orange hair, parted bangs, yellow sports bra, yellow track pants
Huntress: huntress, black eyes, mask, cloak, leotard, navel cutout, purple belt, utility belt, elbow gloves, thigh boots, thigh holster
Supergirl: supergirl, blonde hair, swept bangs, light blue eyes, no pupils, hairband, white t-shirt, cape, miniskirt, gloves, boots
Wonder Woman: wonder woman, blue eyes, constricted pupils, tiara, star earrings, star /(symbol/), strapless leotard, lasso, high heel boots
Vixen: vixen, pixie cut, orange eyes, dark skin, tooth necklace, pendant, tooth earrings, tooth belt, orange bodysuit, cleavage, plunging neckline, belt, elbow gloves, claws
Zatanna: zatanna zatara, black hair, long hair, large breasts, magician, tuxedo, suit jacket, white bowtie, vest, leotard, pantyhose, sleeve cuffs, top hat, high heels
Characters with 5-9 images. Might work with low accuracy:
Harley Quinn (might work because the model already knows her): harley quinn, jester costume, domino mask, blue eyes, black lips, makeup, jester cap, cuffs, two-tone bodysuit
Harleen Quinzel (out of costume): harleen quinzel, short twintails, blonde hair, blue eyes, no pupils, black lips
Katma Tui: katma tui, black hair, gray eyes, no pupils, orange skin, medium hair, forked eyebrows
Lois Lane: lois lane, black hair, swept bangs, purple eyes, earrings, suit jacket, purple jacket, pleated skirt
Poison Ivy: poison ivy, red hair, green skin (or white skin), green eyes, strapless leotard, elbow gloves
Roulette: roulette, black hair, hair bun, hair stick, aqua eyes, dangle earrings, sunglasses, looking over eyewear, tattoo, cleavage cutout, sleeveless dress, elbow gloves, fingerless gloves, side slit
Stargirl: stargirl, blonde hair, blue eyes, star print, mask, long sleeves, midriff, red belt, bike shorts, gloves, holding staff
Volcana: volcana, orange hair, very long hair, spiky hair, red eyes, cleavage cutout, long sleeves, midriff, armband, red gloves
Other characters represented by 5 or fewer pictures in the dataset. Likely low accuracy!
Other characters represented by very few pictures in the dataset: Batman, Bruce Wayne, Circe, Flash, Giganta, John Stewart, Lex Luthor, Livewire, Martian Manhunter, Mercy Graves, Princess Audrey, Superman.
Concepts that were trained somewhat and might work:
pointy breasts: Those classic Bruce Timm pointy breasts, though the TV show did not often depict them. Doesn't always work on Pony; however, Illustrious already knows this tag and "perky breasts".
gotham city: Use the tags "gotham city, red sky, dark" for the red sky look. If you use "night", it may blend the red sky with blue.
space station interior, watchtower
green lantern bodysuit
gold bikini, pelvic curtain, neck ring: the gold space bikini from the episode "Hearts and Minds"
(If you need the old v1 tags, they're here.)
Description
Trained longer and with more images. Should have a stronger style and more variety. Recommended strength: 0.8 to 1.0. Use with 2D checkpoints for more accurate style. See the main description for character tags.
314 images x 5 repeats x 8 epochs = 12,560 total
Optimizer: prodigy cosine, learning rate 1.0
min SNR gamma=6
noise offset=0.03
dim=32
FAQ
Comments (8)
Hey, let me know what you think of the new version 3 I just released. I actually think some of the major characters were learned less well than v2, despite the fact that I added more images of each and increased training time. Maybe it's because I added even more random characters for variety. Or maybe my tag dropping strategy (see what I wrote in the description) didn't work. Anyway, I released it anyway because I think the general style and flexibility is stronger.
One of my favorite styles on here. This is great
How come in automatic 1111, I tried using this Lora regardless of Checkpoint model and it doesn't give me an image in the style of the Justice League TV show and instead gives me an anime style image?
Using A1111 probably should not be the cause of the problem. There's several reasons the style might not come through. First make sure you're using a pony model. However, some highly trained mixed models might overwhelm the lora with their own style. (Start with a model like autismmix, luminaverse, or prefect pony.) Using other loras might also overpower it. Prompting for an anime character name could also push the style toward anime. Increasing the lora weight might help, if it doesn't fry the image.
@navelgaze I use the SD 1.5 version and not the pony version. That is the reason the images are not turning up as I hope they would.
@boothero24 Oh yeah, the 1.5 version was just something I threw together quickly because a commenter requested it. It might not work well. Try it with different anime or 2d art checkpoints until you find something that works ok, then try upscaling and refining the image. You could also try mixing it at lower weight with other character or style loras that complement it.
How do I give the characters different skin tones? For example, I used African American or dark skin and it was supposed to give them brown skin but it still look like a darkened Caucasian skin tone. I was hoping that the characters have more variety in skin tones.
The darkest skin toned characters in the show were john stewart, vixen, and others with similar skin tone, so it may be hard to get this to generate darker than that. The tag "dark skin" was trained and pony models should understand it, though I recall you said you were using the 1.5 version. If that doesn't work, you can try manually re-coloring your images and doing an img2img over them, or use an additional lora specifically for skin tone.