Make sure to use UI that supports v-prediction, otherwise you will get a noisy blob!<h3 id="main-parameters:">Main parameters:</h3><ul><li>Steps: 28</li><li>CFG: 3-5</li><li>Sampler-scheduler: euler/euler a-normal, DDPM-exponential (better noses)</li><li>CFGZeroStar and sometimes Self-Attention Guidance: default for difficult backgrounds</li></ul>Try to BREAK prompts into parts by 75 tokens max, this goes for most models..<h3 id="positive-prompt:">Positive prompt:</h3><pre><code>number of characters (1girl/1boy), character name, artist prompt, scene/environment/camera angle, action, expression, items, quality prompts (masterpiece, best quality)</code></pre><ul><li>some heavily stylized characters might require prompt: nose, realistic.. you can put nose into Adetailer</li></ul><h3 id="negative-prompt:">Negative prompt:</h3><pre><code>lowres, worst quality</code></pre><ul><li>even longer negative should work, but I prefer simplicity.. tags worth noting include sketch, abstract, url</li></ul>Enjoy..

Grim Noire vPred

Here I finally realized, that I need to either add a lot of backgrounds or use LoRA for it.. Used <a rel="ugc" href="https://civitai.com/models/1061443/versatile-backgrounds-illustrious">Versatile Backgrounds Illustrious</a> at 0.4 and <a rel="ugc" href="https://civitai.com/models/510261?modelVersionId=567119">SPO-SDXL_4k-p_10ep_LoRA_webui</a> at 0.6 with combination of my custom lora of 372 images at 0.9.. I have also given up on trying to fiddle with weight blocks, for general use simple approach seems to yield better results AND I DONT SPEND A WHOLE DAY TESTING OPTIONS even though I kinda know what my lora does on various blocks..

Files with SHA256: 95473fd28b076528de8dee1e07c0e9d42a3964d4d4066f7e1060bc18dfc58b7a