Based on <a target="_blank" rel="ugc" href="https://civarchive.com/models/795765/illustrious-xl">Illustrious-XL</a> and <a target="_blank" rel="ugc" href="https://civarchive.com/models/833294/noobai-xl-nai-xl">NoobAi</a> so please visit that page for a rundown of prompts, tags and other baked in features. This model is a merge of Styles I baked and then mixed on NoobAi-Epsilon-1.1. <h3 id="my-generation-settings-for-v1-2:-bk4ffqr60">My Generation Settings for V1-2:</h3><ul><li>Sampler: Euler A AYS</li><li>Sampling Steps: 24</li><li>Upscaler settings:<ul><li><a target="_blank" rel="ugc" href="https://openmodeldb.info/models/4x-UltraSharp">4x-UltraSharp</a> or <a target="_blank" rel="ugc" href="https://openmodeldb.info/models/4x-AnimeSharp">4x-AnimeSharp</a></li><li>1.5x - 2.0x Upscale</li><li>10 Hires steps</li><li>Denoise 0.1-0.6 (Use lower denoise on higher Upscale)</li></ul></li><li>Suggested Dimensions: 832x1216, 1216x832, or 1024x1024</li><li>CFG Scale: 4-6 </li><li><a target="_blank" rel="ugc" href="https://github.com/Bing-su/adetailer">ADetailer </a>along with<a target="_blank" rel="ugc" href="https://github.com/Bing-su/adetailer"> </a><a target="_blank" rel="ugc" href="https://civarchive.com/models/293448?modelVersionId=331333">Anzhc's Face Seg Prototype</a><ul><li>Inpaint Denoise: 0.2-0.5</li><li>Use Seperate width/height:<ul><li>1024x1024</li></ul></li><li>Inpaint mask blur: 24</li><li>Inpaint only masked padding, pixels: 48 </li></ul></li><li>Positive prompt base:<ul><li>masterpiece, best quality,</li></ul></li><li>Negative prompt base:<ul><li>worst quality, low quality, bad anatomy, watermark </li></ul></li></ul>Thank you: @mfcg for nicely baked styles used in the mix. @bluvoll for the assist with training style loras that ended up being made for this merge.Important note: Baked in characters for illustrious/NoobAi are often burnt so you may need to reduce weight so it doesn't make the ckpt look strange.This merge was me wanting to get a nicer cleaner looking style using NoobAi-1.1 as a base. It came out very close to what I want but I need to explore more for data to add in the future especially for the background. I hope you all enjoy. Check out the samples above for inspiration and/or ideas.

Veridium

V2: Remixed with less loras and a different merging method.Goal was to cleanup the base gens of V1 which was too mushy and lost detail. V2 has fixed that now and the images are a lot cleaner, more detailed and sharper while still keeping mostly the same style as V1.

Files with SHA256: 64ef3e241a4adad0b9b785bdb67e25157694daa4ef7a7a23ffcc747261c3c107