Make sure to use UI that supports v-prediction, otherwise you will get a noisy blob!See about version for credits!<h3 id="main-parameters:">Main parameters:</h3><ul><li>Steps: 28</li><li>CFG: 4</li><li>Sampler: euler a (euler for v6 and lower)</li><li>Scheduler: normal</li><li>Im also using CFGZeroStar and sometimes Self-Attention Guidance: default on difficult backgrounds)</li></ul><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: realistic</li><li>you can also try to put nose into faceDetailer</li></ul><h3 id="negative-prompt:">Negative prompt:</h3><pre><code>lowres, worst quality</code></pre><ul><li>url against watermark</li><li>lineart, sketch, dot nose for realism</li></ul>Enjoy..

[BPM] Blink's paradise merge (v-pred)

Started fresh, this is a more bare-bone version. Used only comfyui for merging everything at the same time. Less seems to be more sometimes.Models used, go like them!<ol><li><a target="_blank" rel="ugc" href="https://civitai.com/models/520661/2dn">https://civitai.com/models/520661/2dn</a></li><li><a target="_blank" rel="ugc" href="https://civitai.com/models/833294?modelVersionId=1190596">https://civitai.com/models/833294?modelVersionId=1190596</a></li><li><a target="_blank" rel="ugc" href="https://civitai.com/models/597695/smooth-mix-lora-illustriousponyflux">https://civitai.com/models/597695/smooth-mix-lora-illustriousponyflux</a></li><li><a target="_blank" rel="ugc" href="https://civitai.com/models/169170/xl-outline-ponysdxlkohaku">https://civitai.com/models/169170/xl-outline-ponysdxlkohaku</a></li></ol>

BPMBlinksParadiseMerge_v2.safetensors

Prototype

This is a first try, so don't expect too much. Wanted to keep it bloody simple, I like dark theme.<ul><li>Steps: 28 (more should give slightly better quality, but loong gen time)</li><li>CFG: 4</li><li>Sampler: euler</li><li>Scheduler: normal</li><li>Im also using CFGZeroStar, RescaleCFG: 0.3, Self-Attention Guidance: default)</li></ul>Upscaling helps a lot and I couldn't live without inpainting. :3That's what i use so far, you can try what works for you the best..Positive prompt:number of characters (1girl/1boy), character name, artist prompt, scene/environment/camera angle, action, expression, items, quality prompts (masterpiece, best quality)Negative prompt:lowres, worst qualityMake sure to use UI that supports v-prediction, otherwise you will get a noisy blob!v2:Started fresh, this is a more bare-bone version. (see about version for credits)v1:Started fresh, with similar base, but different weights, i liked it more, merged a few LoRAs with base NOOBAI-XL v-pred and then with 2DN checkpoint. (see about version for credits)Seems to be better overall, keeping the dark esthetic.Prototype:Merged a few LoRAs with base NOOBAI-XL v-pred. (see about version for credits)Enjoy!

Files with SHA256: 442825c5f4ce4c39a9a26e0ae27e0f1789f52581002b604667ac55e5f5d75af5