R-18 previev of a model is here: <a target="_blank" rel="ugc" href="https://www.pixiv.net/en/artworks/113200172">https://www.pixiv.net/en/artworks/113200172</a><h1 id="heading-1267">V2 info:</h1>Model of Boss and comrade from a game Vindictus (Mabinogi Heros) - Claire.Model was trained on 77 images, with 10 repeats in 8 epochs. 3080 steps in total.Model was trained with 2 Checkpoints: Lyriel and CitrineDreamMix. Main one is CitrineDreamMix.Model contains a 3 outfits from her Comrade (no helmet for a armor, sorry).Activation tag: Claire_(vindictus)Appearance prompts: long hair, purple hair, purple eyes, hair between eyes, lipsOutfit 1 (armor) prompts: armor, shoulder armor, gauntlets, breastplate, knight, full armor, chainmail, plate armor, boobplate, black cloakOutfit 2 (officer) prompts: skirt, high heels, knee boots, gloves, hat, cleavage, boots, belt, armor, thigh strap, cleavage cutoutOutfit 3 (sexy) prompts: hair ornament, dress, cleavage, jewelry, thighs, legs, pelvic curtain, thighs, detached sleeves, panties, elbow gloves, feathersHelpfull prompts: 3D, large breasts/medium breasts (you can change their size), realisticFor other prompts, check the images on post.Reccomended weight is 0.6-0.8. On 1 may appear something from her training images, especially everything will become wooden. Plus, sharpness can become super high.Model is trained on 1024x1024 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.Better to use a 3D models, 'cause she was training only on in-game screenshots, and realistic arts.For unclothing, use a tags: NSFW (will change her at all), nude, nipples, off-shoulders. She trained for some poses, so you can try many of them.Even after a many trainings, she still have some problems with a back view. In training, there was not so many images from behind. Just for example. Really, better use a armor. Why to watch on her butt? She's not for ecchi.<h1 id="heading-1268">V1 info:</h1>The version, that was to make a images for training V2. Sometimes it can be better, but it's more bounded to a "wooden" background.

Files with SHA256: 0b3f6bef42469b6d94b004f14a29c509278bc86b67455c5b2b09c19fb4ca5194