Model of boss and comrade from a game Vindictus (Mabinogi Heros) - Arcana.
V1 is more similar to a character + much more beautifull, but not so good at working with a backgrounds and some poses. V2 got it the other way around. You can use a V2 to make a image, then use a V1 for img2img. But it's not always neccesary.
V1 Model was trained on 64 images, with 5 repeats in 10 epochs. 1070 steps in total.
V2 Model was trained on 120 images, with 5 repeats in 10 epochs. 2000 steps in total
Model was trained with Checkpoint: CitrineDreamMix.
Activation tag: Arcana,
Appearance prompts: long hair, purple eyes, purple hair, earrings, mole, mole under eye, large breasts, nail polish,
Outfit 1 prompts: cleavage, bare shoulders, japanese clothes, kimono, fishnets, choker, goggles, goggles on head, bracelet, gloves, fingerless gloves, detached sleeves, white thighhighs, underwear,
Outfit 2 prompts: thighhighs, cleavage, fake horns, horned tiara, high heels, fishnets, detached sleeves, demon girl,
Helpfull prompts: 3D, realistic, jewelry,
For other prompts, check the images on post.
Reccomended weight is 0.6-0.8.
Model is trained on 512x512 resolution, so it's better to use 512x512, 768x512 and 512x768 resolutions to make a images, then use img2img with denoising 0.5-0.55 to make it x2 bigger.
For unclothing use a tags: NSFW (will change her at all), nude, completely nude, nipples. Don't forget to unable the clothes prompts!