<h1 id="vrchat-furries">VRChat Furries</h1>Discord server to see or share other generated images: <a target="_blank" rel="ugc" href="https://discord.gg/bC6U4GwbAt">Discord Link</a><h2 id="new-version--v42">New version : V4.2</h2>Trained for the Yiffymix v3.3 model (<a target="_blank" rel="ugc" href="https://civarchive.com/models/3671?modelVersionId=143769">https://civarchive.com/models/3671?modelVersionId=143769</a>)YOU NEED TO REDUCE THE STRENGTH OF THE MODEL TO &lt;0.8<ul><li>Work better with CFG Scale between 3 and 6</li><li>Works with clip skip 1 and 2</li></ul>Contains :<ul><li>Rexouium (keyword: <code>rexouium</code>)</li><li>Avali with kitavali and davali (keywords: <code>avali, kitavali</code> or <code>avali, davali</code>)</li><li>Nardoragon (keyword: <code>nardo</code>)</li><li>Deuzear (keyword: <code>deuzear</code>)</li><li>Wickerbeast (keyword: <code>wickerbeast</code>)</li><li>Protogen (keyword: <code>protogen</code>)</li><li>Taidum (keyword: <code>taidum</code>)</li><li>Awtter (keyword: <code>awtter</code>)</li></ul>Also works well with anime boy and anime girl<h2 id="loras-working-well-with-vrcfurry">LoRAs working well with VRCFurry</h2><ul><li>Gacha splash LORA v3.1 and v4 : painting style + epic background and arrangements + improves faces (biased towards shorts + thinner body + centered subject) <a target="_blank" rel="ugc" href="https://civarchive.com/models/13090?modelVersionId=26536">https://civarchive.com/models/13090?modelVersionId=26536</a></li><li>HD Helper : avoid duplicate faces on high resolution pictures (<a target="_blank" rel="ugc" href="https://civarchive.com/models/110071?modelVersionId=118644">https://civarchive.com/models/110071?modelVersionId=118644</a>)</li><li>Fluffyrock Quality Tags : slightly improve characters + more likely to generate furry instead of humans (<a target="_blank" rel="ugc" href="https://civarchive.com/models/127533?modelVersionId=151790">https://civarchive.com/models/127533?modelVersionId=151790</a>)</li><li>Micro Waist : self descriptive (<a target="_blank" rel="ugc" href="https://civarchive.com/models/4602/waist-slider-microwaist">https://civarchive.com/models/4602/waist-slider-microwaist</a>)</li></ul><h2 id="negative-embedding-working-well-with-vrcfurry">Negative Embedding working well with VRCFurry</h2>I tried a bunch of different negative embedding and those are the ones that seems to give the best results<ul><li>Boring_e621 (<a target="_blank" rel="ugc" href="https://civarchive.com/models/87781/boringe621-negative-embedding-improve-image-quality">https://civarchive.com/models/87781/boringe621-negative-embedding-improve-image-quality</a>)</li><li>veryBadImageNegative (<a target="_blank" rel="ugc" href="https://civarchive.com/models/11772/verybadimagenegative">https://civarchive.com/models/11772/verybadimagenegative</a>)</li><li>BadDream (<a target="_blank" rel="ugc" href="https://civarchive.com/models/72437?modelVersionId=77169">https://civarchive.com/models/72437?modelVersionId=77169</a>)</li><li>EasyNegative (<a target="_blank" rel="ugc" href="https://civarchive.com/models/7808?modelVersionId=9208">https://civarchive.com/models/7808?modelVersionId=9208</a>)</li></ul>I recommend to use all of them à the same time.<h2 id="keywords">Keywords</h2>Improvers :<ul><li><code>good lights</code> : makes volumes pops more</li><li><code>good colors</code> : better color choices</li><li><code>good composition</code> : more assertive composition choices</li><li><code>good face</code> : faces I prefer</li><li><code>bestart</code> : used this keyword on the best art of my dataset (few example so it is pretty weak/random)</li><li><code>aestheticcartoon</code> : more assertive colors and texture choices</li><li><code>aestheticmuscles</code> : targets no fat, no steroid or bara look (sometimes the position gets a bit stiff)</li><li><code>musculegs</code> : to counter act the strong biasis of most models toward thick upper body and thin legs</li></ul>Worsener : (to use in the negative prompt)<ul><li><code>bad face, ugly, dumb face</code> : used in dataset, pretty self descriptive</li><li><code>bad hands</code> : used in dataset, pretty self descriptive</li><li><code>bad muscles</code> : unaesthetic muscle insertions, poor definition or disproportionned</li><li><code>bad abs</code> : same but specific to abs</li><li><code>ghbelly</code> : used in the dataset for muscular bodies with gh/steroid bubble gut, and distended abs</li><li><code>thin legs</code> : self descriptive</li><li><code>small eyes</code> : self descriptive (but pretty weak as there are very few examples in the dataset)</li></ul>Non vrcfurry keywords but that I like to use with this model<ul><li><code>gorgeous eyes</code> : improve eyes</li><li><code>extremely detailed furs</code>, <code>4k</code>, <code>8k</code> : add details but often biased towards small eyes</li></ul>Styles (keywords) :<ul><li><code>comic</code></li><li><code>thick lines</code></li><li><code>photography</code></li><li><code>render</code></li><li><code>painting</code></li><li><code>swaggy</code></li><li><code>draft</code></li><li><code>hardstyle</code></li><li><code>detailed style</code></li></ul><h3 id="composition-keywords">Composition (keywords):</h3><ul><li><code>rim light</code></li><li><code>dynamic scene</code></li><li><code>complex scene</code></li><li><code>bicolor</code></li><li><code>monocolor</code></li><li><code>morning</code></li></ul><h3 id="body--face-keywords">Body / Face (keywords):</h3><ul><li><code>cute</code></li><li><code>teenager</code></li><li><code>( - / slightly/ very / extremely) muscular</code></li><li><code>ripped muscles</code></li><li><code>(spiky / wavy / feather) hairs</code></li><li><code>relaxed</code></li><li><code>vore</code></li></ul><h3 id="clothes-keywords">Clothes (keywords):</h3><ul><li><code>( - / tight) shirt</code></li><li><code>short</code></li><li><code>legging</code></li><li><code>hoodie</code></li><li><code>bandana</code></li><li><code>croppedhoodie</code></li><li><code>muscles through clothes</code></li></ul><h3 id="position-keywords">Position (keywords):</h3><ul><li><code>side view</code></li><li><code>3/4 view, three-quarter view</code></li><li><code>back view</code></li><li><code>looking back</code></li><li><code>looking down</code></li><li><code>looking away</code></li><li><code>defying look</code></li><li><code>dynamic position</code></li><li><code>seated</code> / <code>sitting</code></li><li><code>crouching</code></li><li><code>lying on</code></li><li><code>leaning on</code></li><li><code>facing down</code></li></ul><h3 id="suggestive--nsfw-keywords">Suggestive / NSFW (keywords):</h3><ul><li><code>dick through clothes</code></li><li><code>dick sticking out</code></li><li><code>autosucking</code></li><li><code>jacko pose</code></li><li><code>grabbing pec</code></li><li><code>first person view</code></li></ul><h3 id="others-keywords">Others (keywords):</h3><ul><li><code>hmoma</code> : furry with human (at least 2 persons)</li></ul><h2 id="tips--tricks">Tips &amp; tricks:</h2><ul><li>Improve avali look by adding those keywords: <code>cute, big eyes, feathers, black eyes</code></li><li>Get better faces by adding these keywords: <code>close up portrait</code></li><li>Sometimes you can get better faces by changing hairs / furs colors</li></ul><h3 id="good-positive-prompt-keywords">Good positive prompt keywords :</h3>Kind of a meta I found across all my models<pre><code>good lights, good colors, good composition, monocolor, bicolor, good face, (cute:0.7), teenager, </code></pre><ul><li><code>good lights, good colors, good composition,</code> are keywords I added on the best artworks in the dataset</li><li><code>monocolor, bicolor,</code> are often associated to higher quality / artistic artworks in the dataset with interesting color management</li><li><code>(cute:0.7), teenager,</code> cute put emphasis on eyes and make them much better and teenager couteract the "chibi"/"kid" look it could create while still keeping a cute look</li></ul>and some bonus keywords that can make it look even better, even if it depends a lot on your tastes ^^<pre><code>masterpiece, extremely detailed furs, 8k, soft fur, perfect composition, detailed shading, dynamic pattern with accents, detailed style, soft,</code></pre><h3 id="good-negative-prompt-keywords">Good negative prompt keywords :</h3><pre><code>(neck fluff, fluffy neck,:1.2), lores, bad anatomy, warped, bad hands, BadDream, easynegative, boring_e621_v4, (ghbelly, bad face, ugly, dumb face, thin legs, flat hairs, bad abs, fluffy neck, neck fluff, fluff on the neck:1.2), (anime boy, human, human ear:1.4), verybadimagenegative_v1.3</code></pre>There seems to be a bias towards females, so if you want to generate males, use:<pre><code>female, woman, girl, 1girl,</code></pre>If the shirts are cropped / always lifted, use:<pre><code>crop top, cropped shirt, bra, shirt lift, </code></pre><hr /><h3 id="heading-79">Planned for next version (V5):</h3><ul><li>Better tail control</li><li>Reference sheet addition ?</li><li>better dataset captionning</li><li>I don't really know, don't hesitate to send ideas</li><li>no mamagen please</li></ul>

VRChat Furries

same but better on all, normally, and trained on YiffyMix 35

vrcfurry_v43.safetensors

v4.3

v4.2

v3.2

v3.1

v2.1

nardo v1.0

avali v1.0

rexouium v1.1

v1.0

Files with SHA256: 463a57586f83112b9631d28d8e36febcb6d88b22d9452cc8b075afe141991538