Relisted 0.75 noob epsilon by user request, and the most recent epsilon version, but I generally recommend the newer vpred version <a target="_blank" rel="ugc" href="https://civarchive.com/models/1068866/chadmix-finetune-extract-for-noobai-vpred-10">https://civarchive.com/models/1068866/chadmix-finetune-extract-for-noobai-vpred-10</a>v3: i have taken steps to combat problems with red tint. includes itercomp weights.This started as a rough collection of meme edits, but I noticed some startling effects on quality and poses and put work into making it something more. Yes this will turn characters into Gigachads of varying degree. You can adopt the poses independently of the meme character. It will give you Gigachad as an independent character with your waifu. It's also trained on the more acrobatic memes and will stabilize midair acrobatics and dancing pirouettes. I also think it helps with posing human characters generally but I'm biased in favor of my own product. It will bias gens in favor of more detailed characters and stronger contrast, but otherwise plays nice with artist styles.V2.0 update: cleaned up dataset pretty good, jpeg artifacts negative no longer necessary. Hit a real sweet spot for overall contrast level on the bake. Will only produce this for Noob, don't want to go back to regular Illustrious.How to use:<ol><li>Strongly recommended to use masterpiece in positives, and worst quality and jpeg artifacts in negatives. The model was trained on some very dubious quality images for broader character knowledge and I did my best to mitigate their effect on quality.</li><li>There are three new tags:<ul><li>the chad smile -- associated tags include upper teeth only, smile, grin, portrait. beard, stubble, manly and so on affect degree of chadification</li><li>the stoic face -- looking to the side and serious help in recreating the other meme facial expression</li><li>sitting gigachad style -- arm on thigh, hand on own thigh, own hands together help recreateThese three tags take pretty well to &gt;1 tag power to help get the desired effect</li><li>and gigachad \(meme\) is on the dataset as a whole</li></ul></li><li>I made versions for NoobAI 0.5 and base Illustrious 0.1, but i really recommend Noob.</li></ol>

Gigachad & poses, IllustriousXL & NoobAI

i have taken steps to combat red tintvpred 0.6 + itercomp merge, trained on that, 64dim extracted from vpred 0.6

Files with SHA256: 9aa8f67e90801e022f701a1dd47c0cdb8ddd4b07a00ec6d5ab1a6b559765a41e