<h2 id="heading-839">V3.0</h2>I merged the newest state of next Realism Engine in training coming soon ™If you get artifacts please readold version, you must use a VAE <a target="_blank" rel="ugc" href="https://huggingface.co/stabilityai/sdxl-vae/blob/main/sdxl_vae.safetensors">https://huggingface.co/stabilityai/sdxl-vae/blob/main/sdxl_vae.safetensors</a>UPDATE: For 3.0 I uploaded a version with the VAE baked in and removed the old file.There's some artifacts issues with those 3 samplers. You can either increase the steps count to get rid of them or use another sampler.<ul><li>DPM++ 2M SDE (dpmpp_2m_sde_gpu on comfy and Fooocus) : ok with 60 steps</li><li>DPM++ 2M SDE Heun : ok with 90 steps</li><li>DPM FAST : ok with 60 steps</li></ul>It's usually a good practice to use Karras Sampler when you use Karras scheduler.Here's a list of the best samplers tested that will help to get sharp eyes. (Name can vary on other clients)<ul><li>DPM++ 3M SDE Karras</li><li>Restart</li><li>DMP++ SDE Karras</li><li>DMP++ 2M SDE Karras</li></ul>Happy Prompting<h2 id="heading-839">V2.03</h2>Since the first version was a great success I improved on it to make it even better. In this version I used the same base, I added 2 Lora to improve woman's anatomy and mixed in the current state of the next Realism Engine in training.Happy Prompting<h2 id="heading-840">V1.0</h2>This is a simple merge of 3 models that turned out to be good for full female nudity.Both fp32 and fp16 are available for download.Models used:<a target="_blank" rel="ugc" href="https://civarchive.com/models/130697">foddaxlPhotorealism_v60 (defunct)</a> <a target="_blank" rel="ugc" href="https://civarchive.com/models/159606?modelVersionId=179507">imperfectXLTrueNSFW_v13</a> <a target="_blank" rel="ugc" href="https://civarchive.com/models/152525/realism-engine-sdxl">RealismEngine_SDXL_v1.0</a>Merge formula0.8(0.5(foddaxlPhotorealism_v60) + 0.5(imperfectXLTrueNSFW_v13)) + 0.2(RealismEngine_SDXL_v1.0)

SDXXXL

Improvement on the woman body diversity and anatomy

Files with SHA256: aee3e7f23ee897364e5285f7d9392a9511b4e18e75e471a8f5ab422489ef5f22