<p>Lora for another <strong>OC</strong> I came up with.</p><p></p><p>My OCs collection: <a target="_blank" rel="ugc" href="https://civarchive.com/collections/201272"><strong><span style="color:rgb(250, 176, 5)">here</span></strong></a>.</p><p></p><h3 id="heading-5037"><span style="color:#fab005">Eris</span> is a scout and an experienced hunter. She is part of the warrior caste of her home planet and is well respected within the ranks of the military, providing both active duty service and training new potential soldiers. The harsh environment of her home arid planet have shaped her into a hardened individual.</h3><p></p><h3 id="heading-5037">I will upload two versions - they are variations, difference being the model they were trained on.</h3><h3 id="heading-5037">1) <u><span style="color:#82c91e">SD 1.5 base model</span></u> - this one is extremely versatile, fits very well with almost any model, but lacks certain details from the dataset (i.e. the very dark skin is often not very at all);</h3><h3 id="heading-5037">2) <u><span style="color:#82c91e">URPM 1.3</span></u> - very good detail orientation and faithfulness towards the dataset, but some parts of the images come out blurry and odd. Also there are times freakish bodies are generated (i.e. 3-4 legs, multiple arms, etc.) </h3><ul><li><p>What I could recommend is, try both versions, see what fits better for your idea. Also, what I like to do is use the URPM one to generate the image, and put the SD 1.5 one in the ADetailer prompt to "fix" the face.</p><p></p><p> </p></li></ul><h2 id="heading-5040">Results, as with any model, may vary.</h2><h3 id="heading-2410">I have tested the LoRA, but I obviously cannot produce all possible environments, so this model might work perfectly for your personal settings, negative embeddings and other resources used for image generation <strong><em>or </em></strong>it might be terrible.<br />（づ￣3￣）づ╭❤️～</h3>

PulenKompot-Eris

<p>Trained on SD 1.5 base model (runwayml/stable-diffusion-v1-5)</p>

Files with SHA256: 5de502216cb9ecd3982f3f6f643b49fe7988d809348fe58c229b15c1dd4b5eee