v1 (trained using "Concept") was such a blurry and fuzzy mess that I scrapped it and went to train this v2 (using "Character" instead) lolWould greatly appreciate it if you could read this before posting! <a target="_blank" rel="ugc" href="https://civarchive.com/articles/7060/look-both-ways-before-you-crosspost">https://civarchive.com/articles/7060/look-both-ways-before-you-crosspost</a>Prompts:<pre><code>germanfortifiedsuit</code></pre>Character prompts for the characters used in the training data set (the fourth of which I did not personally test lol):<pre><code>ilfriede von feulner, blonde hair, blue eyes, folded ponytail, single hair bun, sidelocks, large breasts</code></pre><pre><code>lunateresia von wizleben, green hair, yellow eyes, medium hair, swept bangs, large breasts</code></pre><pre><code>helgarose von falkenmayer, blue hair, blue eyes, ponytail, large breasts</code></pre><pre><code>brigitte von wasternach, light brown hair, purple eyes, eyepatch, curtained hair, medium hair, lipstick, large breasts</code></pre>Notes:<ol><li>10 images in the training data set, 7 of which are Ilfriede, so expect her side hair to manifest itself into the images you generate lol</li><li>Trained 20 epoches but ended up going with epoch 5. For some reason the longer it trains, the worse the fingers look lol. As such, clothing detail might not be that accurate. When that happens, adding "black bodysuit" to the prompt tends to help if the front part shows up as silver or gray. Can't help with the chin guard missing though</li></ol>

Files with SHA256: b70a83a0679e24bd3187f1f95861c48738fad18a57254e0417c1a6aec5b43a29