<h2 id="training-information">Training Information</h2>Trained in <a target="_blank" rel="ugc" href="https://github.com/ThetaCursed/Anima-TrainFlow">Anima-TrainFlow</a><ul><li>This LoRA is trained using Anima Base with the total image dataset of 52+ of Arceus in Pokemon.</li></ul><h3 id="promptparameter-settings:">Prompt/Parameter Settings:</h3>Lora Strength: .9 - 1<h2 id="base">Base</h2>Activation Tag:<pre><code>1boy, arceus, furry,</code></pre>Reinforcement Tags:<pre><code>red eyes, green sclera, white body, black body</code></pre><h3 id="after-thoughts">After thoughts</h3><ul><li>The same with <a target="_blank" rel="ugc" href="https://civarchive.com/models/2814513/xxl-delga-monster-ancient-cline-and-mugen-lora-anima">XXL Delga. </a>This lora's dataset has been loosely curated to be leaning more on the anthropomorphic but focuses of being muscular and buff. Also, I added 20+ generated synthetic images using by <a target="_blank" rel="ugc" href="https://civarchive.com/models/1347735/arceus-pokemon">R5050's Arceus IL Lora </a>and <a target="_blank" rel="ugc" href="https://civarchive.com/models/579632/indigo-furry-mix-xl">Indigo Furry Mix XL Chenkin Noob</a></li><li>I'm shocked to find that adding synthetic images with a specific particular body shape seems to help in being more bias in big muscular bodies, for no style tags. However, It a different story if it uses style tag as it can be inconsistent with it output. Sometimes, it will output proportionate anatomy while other it does keeps the big muscular body proportions.</li></ul>

Files with SHA256: 979e4190f780acb59fb692e5e85a60c76bf6e0afa366c80379be08e218e7fc13