<a target="_blank" rel="ugc" href="https://beacons.ai/diffusiondudes"><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/c5d38298-efb3-4fc9-8e8d-cc6f2dec38d8/width=525/c5d38298-efb3-4fc9-8e8d-cc6f2dec38d8.jpeg" /></a><a target="_blank" rel="ugc" href="https://beacons.ai/diffusiondudes">Visit my Patreon/Gumroad/Ko-fi for exclusive access to all 80+ of my LoRAs! ▼▼▼▼▼</a>I'll be releasing 1-2 LoRAs on Civit every month, with early access on my stores, and over 80 exclusive models 💪 Members can get full access to my entire archive plus my private TG for my steamiest content!-Scruffy bums! Fuzzy butts! Unbuzzed buns! Let's face it, no matter what model you use, if you try to get realistic butt hair, the best you get is a tuft. This LoRA aims to create more realisitic rear end hair with more distribution and thicker growth. This is a LoRA trained on the base 1.5 model, at 768px. Please share your results and feedback for any future training improvements.Recommendations:<ul><li><code>Trigger token: scrffybm, hairy ass</code></li><li>Other tags that you can optionally include (these help on some models and make others look messier! Experiment!) - anal hair, anus, male pubic hair, hairy thighs, bara, scrotum, testicles</li><li>Negative prompt: pussy, vagina, vaginal</li><li>The strength is dependent on models, but I recommend anywhere between 0.3-1.2 and experimentation. Look at my grids for details.</li><li>I recommend using LoRA Block Weight extension to control LoRAs on a detailed block-level. For this, OUTALL (1,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1) is a perfect sweet spot that affects less of the image composition. It will totally work fine without block weights though.</li><li>I used a pretty diverse dataset, so customize appearance as you wish</li><li>For upscaling, the effects of this will diminish significantly if you use Hi-res fix above 0.4, I recommend 0.25-0.35. With Tiled Diffusion, it works well around 0.7, with ControlNet Tile on.</li><li>This was only trained on realistic images, and unfortunately it seems most 2D/2.5D/Anime models don't handle it as well as realistic ones do, but they can make it work with a bit of experimentation.</li></ul>I am not responsible for your use of this resource, which is not for commercial use and is for research purposes only.

Files with SHA256: 94ae1ada11da05ff4d8b0cf6ffaa376ba0814ef05b97ebb053ca8ebcb4319444