Bunch of these preview images got marked R cause her outfit Hello! 2nd LoRA is here. This one was not easy. hack//sign is 480p at best and the character shots are very lacking. Halfway decent full body were almost impossible. This one has 55-69 images in the training data. I had to process the screenshots twice. First for cleaning and next for enhancing the cleaned versions. This improved the data by a lot. Ive included 1 set for example and tagged it as Original &gt; Cleaned &gt; Enhanced.<h3 id="v2.0-il">v2.0-IL</h3>This is v2.0 for Illustrious. First time using Civitai to train.. was weird.<hr /><h3 id="v2.0-(anima)">v2.0 (Anima)</h3>IMO this version is better in every way I could test compared to the others. If you want style closer to the anime, dont use high res types of tags.In the preview images i used <a target="_blank" rel="ugc" href="https://civarchive.com/models/2458426/anima?modelVersionId=2945208">AnimaBase v1.0</a> &amp; <a target="_blank" rel="ugc" href="https://civarchive.com/models/2767118/anima-turbo-remix">Anima Turbo Remix v1.0</a>Removed a few poor images and replaced with somewhat better ones. Upscale all as close to 1536 as they could go per aspect ratio Bucketed 512x1536, Main 1536 training DoRA trained TE trained Rank Budget 4-24. Half file size, same quality. Original 139.3mb --&gt; 73.6mbTrigger: mimiru Triggers: black choker, red bikini armor top, silver elbow guards with spike, white groin armor, short black loin cloth, black boots, greatsword, transparent pantyhose Strength: 1.0<hr /><h3 id="v1.5-te-(anima)">v1.5-TE (Anima)</h3>This was more experiment again. Various adjustments, few more training images, TE trained, epoc 10, rank 24, more captions, more triggers. Triggers: mimiru, brown hair, green eyes, red bikini armor top, silver elbow guards with spike, white groin armor, short black loin cloth, black boots, transparent pantyhose Strength: 1.0-1.1<hr /><h3 id="v1-(anima)">v1 (Anima)</h3>First attempt. Triggers: mimiru Strength: 1.0-1.25 Note: Yes, she does wear stockings. No, its not easy to tell in the show. When you see them you never unsee it though. The training picked this up really well and you have to prompt it out if you dont want them.<hr />I do this for fun and to learn. Likes, comments and reactions are nice to see if my efforts are worth while and posting. I dont really care about the buzz or whatever but drop me a comment and like if you enjoy what Ive done. &lt;3

Mimiru - hack//SIGN

Removed a few poor images and replaced with somewhat better ones. Upscale all as close to 1536 as they could go per aspect ratio Bucketed 512x1536, Main 1536 training Rank Budget 4-24 DoRA trained TE trained Half file size, same quality

Files with SHA256: 0ddc6d54191574005a7e8bbb56fb11e2dc574439ff56fbc8cc5623169c61d263