AAM - AnyLoRA Anime Mix

CHECKPOINT

Original

Updated: Nov 26, 2023 5:35 AM

cgmech,1girl, weapon, long hair, breasts, leotard, solo, gun, holding, thighhighs, holding weapon, very long hair, garter straps, holding gun, ass visible through thighs, animal ears, bare shoulders, looking at viewer, white hair, mechanical arms, black leotard, standing, outdoors, black thighhighs, cowboy shot, elbow gloves, cleavage, medium breasts, single mechanical arm, gloves, parted lips, dual wielding, strapless, highleg leotard, legs apart, building, arms at sides, cable, lips, day, large breasts, cat ears, science fiction, highleg, covered navel, detached collar, strapless leotard, handgun, arm at side, bangs, yellow eyes, fake animal ears, black gloves, cyborg, eyepatch, parted bangs, backlighting, headgear, global illumination, hdr, ((film grain, blurry background, blurry, glow, natural light, intense sunlight)), scifi, glow, radiosity, ((masterpiece, best quality)), <lora:cgmechmix_offset:1><lora:more_details:0.3>

cgmech, (face mask, menpo), upper body, underboob, portrait, robot,white orange armor, white shimmering hair, neon light, 8K, RAW, best quality, masterpiece, ultra high res, colorful, (medium wide shot), (dynamic perspective), sharp focus , (depth of field, bokeh:1.3), extremely detailed eyes and face, beautiful detailed eyes,large breasts,(black gold, trimmed gear:1.2),(In a futuristic weapons factory:1.2), ((masterpiece, best quality)), <lora:cgmechmix_offset:1><lora:more_details:0.3> Detailed background, spaceship interior <lora:Niji:0.5>

masterpiece, best quality, hatsune miku, white gown, angel, angel wings, golden halo, dark background, upper body, (closed mouth:1.2), looking at viewer, arms behind back, blue theme, cloud

megumin, 1girl, bare shoulders, black cape, black gloves, black hair, blush, cape, choker, collarbone, dress, fingerless gloves, gloves, hair between eyes, hat, long sleeves, looking at viewer, medium hair, off-shoulder dress, off shoulder, red dress, red eyes, sidelocks, simple background, solo, indoors, witch hat, ((masterpiece))

8k portrait of beautiful cyborg with brown hair, intricate, elegant, highly detailed, majestic, digital photography, art by artgerm and ruan jia and greg rutkowski surreal painting gold butterfly filigree, broken glass, (masterpiece, sidelighting, finely detailed beautiful eyes: 1.2), hdr,

masterpiece, best quality, hatsune miku, white gown, angel, angel wings, golden halo, dark background, upper body, (closed mouth:1.2), looking at viewer, arms behind back, blue theme, stars, starry night

(red and black mecha), gundam, holding rifle, explosion, laser, robot, masterpiece,best quality,ultra-detailed,very detailed illustrations,extremely detailed,intricate details,highres,super complex details,extremely detailed,cowboy shot, caustics,reflection,ray tracing,demontheme,nebula,dark aura,cyber effect, action, ancient japanese architecture,pond, starry sky,skyline, <lora:mecha_offset:1> <lora:more_details:0.2>

cgmech,multiple view angles, weapon, 1girl, white hair, breasts, high heels, navel, bodysuit, gun, black hair, elbow gloves, short hair, gloves, realistic, medium breasts, knife, science fiction, boots, holding, simple background(photorealistic, film grain:1.3), ((masterpiece, best quality)), <lora:cgmechmix_offset:1><lora:more_details:0.3> simple background

AnyLoRA Anime Mix (AAM)

Anime Screencap Style Model

Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee ☕. A ❤️, a kind comment or a review is greatly appreciated.

Purpose of this model

Make amazing anime style artwork on its own.
Train character loras where the dataset is mostly made of anime screencaps, allowing less anime style transfer and less overfitting.
Use anime styles if the lora is "weak" and doesn't give enough flat coloring.
It's also pretty good for generating hentai stuff.

Suggested settings

Set the ETA Noise Seed Delta (ENSD) to 31337
Set CLIP Skip to 2
DISABLE face restore. It's terrible, never use it
Use negative prompts and embeddings that don't ruin the style
Use AnimeVideo or Foolhardy as upscalers. If you know what to do, Latent makes an interesting effect (see the first example and the Miku one)
(optional to reproduce some results) Set "Do not make DPM++ SDE deterministic across different batch sizes." Only my old images use that. (this setting is not actually good, so I'm gonna disable it, but might be the reason some results are different)

Brief history

Chatting on Discord, I noticed my Mushoku Tensei style LoRA, while trained on AnyLoRA, was performing a bit better on CyberAlchemist's AnimeMix base model. I read on his HF that the model was a supermerge of AnythingElse, his AnimeMix lora and various other style loras. Since AnimeMix was already including many of my old anime style loras, I wanted to experiment and try to make a model that could maximize the result of my Mushoku Tensei style, going beyond what AnyLoRA and AnimeMix could do.

After testing many different versions, adjusting the settings and receiving feedback from the Discord channel, I ended up with a version that works very well on anime loras trained on AnyLoRA (base). Especially my Mushoku Tensei one which felt very weak on AnyLoRA.

I don't advise using this for anime style training (limit to generation after training it on AnyLoRA or NAI), but it might actually be fairly good for character loras where your dataset is made mostly of anime screencaps. It should technically reduce the anime style impact making the character lora less overfitted. I'll definitely experiment with it in the following weeks.