<h3 id="description-sbx1ynagz">Description</h3>FurryGoNuts XL is the continuation of my model series aimed at delivering high quality furry artwork. It tends to lean more realistic than anime/kemono when left to defaults (as that's where my tastes lie), but is very capable of generating many different art styles.Recommended SettingsSampler: Use your favoriteCFG: 7+Steps: 30+Version 1.3bImprovements to overall image quality and prompt recognition/adherence. I noticed that v1.2 wasn't as stable as I wanted it to be, so that has also been significantly improved.Version 1.2This version has had a large jump in general image quality and prompt adherence. Much improved species recognition, as well.Version 1.1b (Test Release)Incremental quality update with better Pony prompt adherence. This release is still under testing.Version 1.0-RCThis is the first "official" version of the model. I've improved significantly on the shortcomings of the Pre-Release model.<ul><li>Hands and feet generation have been improved, but could always use more improvement.</li><li>NSFW performance is definitely better, especially where male subjects are concerned.</li><li>General image quality has received an improvement, but it may be harder to get photorealistic results.</li><li>Signatures, watermarks, and other unwanted artifacts should show up less now, and be more responsive when using negative prompts to remove them.</li></ul>Initial Pre-Release (0.1h)The first release of my FurryGoNuts XL model. It's still very much a work in progress, so please try to leave constructive criticisms and I'll do my best to resolve the issues in future releases.Known Issues<ul><li><s>Poor NSFW Performance</s><ul><li><s>This model is currently known to struggle with genitalia (particularly male), and is known to have issues generating multiple individuals in sexual congress. Correcting this behavior is my top priority for this model.</s></li></ul></li><li><s>General Model Stability</s><ul><li><s>I've noticed that, on rare occasions, using too many seemingly contradictory prompts (or having a few with relatively high strength of &gt;1.3) can cause the model to generate a blank image or simply noise. I'm tracking down the cause.</s></li></ul></li><li><s>Poor Hands/Feet</s><ul><li><s>As this model progresses, I'll be concentrating on improving the hand and foot generations.</s></li></ul></li><li><s>General Image Quality</s><ul><li><s>While I believe the image generation quality of this model is acceptable, or at least enough for a pre-release candidate model, I'm going to concentrate on improving it along with posturing and other image issues.</s></li></ul></li><li><s>Signatures/Watermarks</s><ul><li><s>I feel like it generates a disproportionate number of these artifacts, even when strongly in the negatives. I'm guessing it's from poisoned training data, but this is a merge, so it can't be helped. Likely it'll iron itself out as the model is fine-tuned.</s></li></ul></li></ul>Unknown Issues🤷‍♂️

FurryGoNuts XL

Further improvements to quality and prompt recognition/adherence.

furrygonutsXL_v13b.safetensors

Files with SHA256: 991ff7b82e60b5cda7e426a0bbd94ff6785c57fb6a0bd856b18eb372bf3dea2a