Trained her so many times I lost count, but close enough. Team Yen FTW! So v1 was a very early try, v2 is the ~1.000.000th try. -it's called dedication-v4 is mostly not from the game but there is some, it has a basic understanding of both her outfits but not perfect, doesn't even matter I would say. Happy how this one turned out! Might not need to do another one.v34-2 I don't think you can get any more accurate than this on IL, exact same data set as v34, but different setting (I did process the images with varios methods but still, the same set of images content wise). For some strange reason tho IL did that thing where you need the negative "child" otherwise she has child like features, but maybe thats a checkpoint thing, or setting, I dunno, maybe... - No promises I won't make any more versions of her!v34 has no outfit, it was trained on nudes from various sources that seemed very close to the original (it knows the choker tho, thats iconic)IL v4 Trained on IL XL 1.1 - the rest is IL v0.1<hr />Recommended style tags:<pre><code>3d \(artwork\), realistic</code></pre><hr />v4 Outfits:<ol><li>choker, dress, feather trim, arm wraps, gloves, loose belt, thighhighs, thigh boots</li><li>choker, collared shirt, jacket, fur trim, layered sleeves, belt, black gloves, black pants, thigh boots, belt boots</li></ol>v2 Outfit: choker, pedant, jacket, white shirt, black gloves, fur trim, belt, pants, thigh boots, belt boots, string braceletv1 Outfit: choker, dress, gloves, corset, off shoulder, stockings, thigh boots, high heels<hr />I always generate with Euler a, 25 steps, CFG 5, very basic for IL and pony.Schedule type: automatic I use No embeddings, NO negatives and NO quality tags at all! Only the style and medium tags.Recommended Resolutions: (SDXL standards) - Pony/IL too640 x 1536 / 1536 x 640768 x 1344 / 1344 x 768832 x 1216 / 1216 x 832896 x 1152 / 1152 x 8961024 x 1024

Files with SHA256: 75f0ca2331a697a0528efb4dfda4b454753d4c85aeaffc78673bdc014df8551e