My attempt at <a target="_blank" rel="ugc" href="https://civarchive.com/bounties/2328/san-mononoke-hime">this </a>bounty.<h2 id="comments-zb2hz5r1k">Comments</h2><ul><li>To have her wear her mask, remove tags relating to her eyes, face paint, etc., and try adding "mask on face" or "mask" to the positive prompt. Prompting it can be a bit tricky as the dataset contained ~6 examples of her wearing the mask on her face. The training set did include an additional 6 photos of cosplay masks.</li><li>Every image was manually labeled. (Invalid tags were used in training for Pony V6 XL and Animagine XL V3.1.)</li><li>The model was not trained on explicit material.</li><li>Most of the training data contains her core outfit, and may sometimes require a bit of tinkering to change to a novel outfit.</li></ul><h2 id="images-424nhbniw">Images</h2><ul><li>Common training sizes: 1280x704 (x16), 1344x768 (x13), 832x1152 (x10)</li><li>Signatures were removed from fan art for training.</li><li>x114 images of San<ul><li>x32 screenshots (x2 repeats)</li><li>x62 fan art</li><li>x14 cosplay</li><li>x4 pvc statue</li><li>x2 yarn/cloth doll</li></ul></li><li>x6 photos of isolated cosplay masks</li></ul>

San (Princess Mononoke)

<ul><li>Common training sizes: 1280x704 (x16), 1344x768 (x13), 832x1152 (x10)</li><li>x114 images of San</li><li>x6 photos of the isolated cosplay mask</li><li>Trained on PonyV6</li><li>Rank 8, Alpha 4</li><li>Trained on PonyV6.</li><li>Testing renders made with AutismMix.</li></ul>

Files with SHA256: 66512e48eca4acfe1dc810acbba393a860f8c33402befe627831424aa9dc2b69