<h1 id="hoop-earrings-for-furries-(for-noobaiilqwen)">Hoop earrings for furries (for NoobAI/IL/Qwen)</h1>Problem: NoobAi based models often misplace hoop earrings. Especially with anthros of ungulate species (deer, goats, etc.) the hoop earrings are placed where they would be for a human character, making the earrings float in the air below the actual ear.Solution: This LoRA. It was trained only on furries to reduce the existing human bias. I don’t claim that the success rate is 100%, it is not, so you might still have to roll the seed a few times for a good result.For 100% success rate I would have to overtrain the concept and I didn’t want that. This LoRA is intended as a small helper resource to “nudge” the model in the right direction and ensure that the impact on the behavior of the base model remains small.Do not use it with several other LoRAs, it will destroy the effect most likely. Most people on Civitai use too many LoRAs at once. Using 7 for one image is never a good idea.Illustrious version:Completely reworked the dataset. Should be much more accurate now than the NoobAI version. Also works with NoobAI, so this version is recommended in almost any case.Aside from that use like NoobAI version.<h2 id="usage">Usage</h2>Triggers are<pre><code>hoop ear ring, jewelry</code></pre>if you have trouble to get good results just look at the showcase images. They all have full metadata.LoRA weight tested up to 2.0 without issues. 1.0 is normally be enough however.Base model used for training: NoobAI-EPS-1.1.<h2></h2><h2 id="qwen-version">Qwen version</h2>Not many people will use it, so I keep it brief: You know how to prompt Qwen I assume. Prompt must contain"hoop earrings" and/or "hoop ear ring".Unstable for some prompts. Then use higher LoRA weight.

Hoop earrings for furries (for NoobAI/IL/Qwen)

Illustrious version:Completely reworked the dataset. Should be much more accurate now than the NoobAI version. Also works with NoobAI, so this version is recommended in almost any case.Aside from that use like NoobAI version.

Files with SHA256: 4bea9e5aa0b40904b91b5def54e3ea42ab848d6e4ce758be811541ff650c53bd