<h3 id="v3-(mxfp8-or-nvfp4-or-int8_convrot-or-q8_qr-or-q5_1-or-q4_cr_w4a4)">V3 (mxfp8 | nvfp4 | Int8_ConvRot | Q8_QR | Q5_1 | Q4_CR_W4A4)</h3><span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">Use this extension's loader for GGUF files: <a target="_blank" rel="ugc" href="https://github.com/molbal/ComfyUI-GGUF">https://github.com/molbal/ComfyUI-GGUF</a>Focused primarily on illustration and painting, with notable NSFW capabilities<h3 id="v2">V2</h3>Model geared more towards illustration<h3 id="v1-why-this-model">V1 - Why this model?</h3><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/54cc3714-a9cb-42de-96b8-658446c555ca/original=true/54cc3714-a9cb-42de-96b8-658446c555ca.jpeg" />I designed Kreation to overcome the limitations of <a target="_blank" rel="ugc" href="https://civarchive.com/models/2729631/kreamania">Kreamania</a>. I could no longer expand that model's graphic and painterly capabilities—which had previously fully satisfied me—without compromising its realism.However, I was unable to achieve specific effects with Kreamania (such as abstract art, Cubism, Symbolism, or stylized drawings). Even when using LoRAs, the shapes remained rigidly tied to realism (for instance, making a square human head was simply impossible).Rather than attempting to modify Kreamania, I chose to explore a new path with this model, prioritizing an unconstrained painterly and graphic style.Kreation is intentionally incapable of producing realistic images; its focus is strictly limited to painterly and artistic styles. This way, both approaches can be developed independently.<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">I'm currently only distributing the optimized inference format (INT8/ConvRot version) to keep performance high and protect the core workflow. The BF16 master weights won't be released publicly for this version. Thanks for understanding!

Kreation

<ul><li>Steps : 8 - 10</li><li>Sampler : heunpp2 / sgm_uniform</li><li>Krea2T Enhancer Advanced preferably disabled</li></ul>

Files with SHA256: a09edda2c00659fd95ffb59413127c0ef758226f30335cd48d6188fcaee5ca9f