This model is based on the Hassaku XL (Illustrious) mix. It is my best creation so far, with a huge potential for creativity. A CFG scale of 3 to 7 is recommended. For the best results, use CFG 6 with the Euler A sampler. Make sure to check out the sample images to see how to get the most out of this model. <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4e02b6a6-ee80-4c25-888f-a8732d630b25/original=true/4e02b6a6-ee80-4c25-888f-a8732d630b25.jpeg" /><h3 id="recommended-settings">Recommended Settings</h3>Base Generation:<ul><li>Sampler: <code>euler_ancestral</code></li><li>Scheduler: <code>beta</code></li><li>CFG: <code>5.5</code></li><li>Steps: <code>25</code></li><li>Size: <code>1024×1344</code></li><li>Post-process: Sharpen (radius 1, sigma 0.20, alpha 0.30)</li></ul>Upscale Stage:<ul><li>Sampler: <code>euler_ancestral</code></li><li>Scheduler: <code>beta</code></li><li>CFG: <code>4.0</code></li><li>Steps: <code>25</code>, start_at_step: <code>20</code></li><li>Post-process: Contrast Fix (blend_factor 0.30, soft_light) + Sharpen (radius 1, sigma 0.30, alpha 0.40)</li></ul> Recommended Prompt: Positive: masterpiece, amazing quality, best quality, highly detailed, intricate details, sharp focus, photorealistic lighting, volumetric lighting, soft lighting, realistic shadows, soft shadows, dramatic shadows, depth of field Negative: bad quality, worst quality, worst detail, sketch, censor, lowres, bad anatomy, low quality, text, watermark, logo, artifacts,<h3 id="recommendation-to-comfyui">Recommendation to ComfyUI</h3>In portrait mode, the eyes are drawn pretty well, but with "distant" character generation, the eyes lead (this is the base), my advice is to use FaceDetailer.FaceDetailer Settings:<ul><li>Guide Size: 1024</li><li>Max Size: 1024</li><li>Steps: 30</li><li>CFG: 3.5</li><li>Sampler: euler</li><li>Scheduler: karras</li><li>Denoise: 0.35</li><li>Feather: 12</li><li>Bbox Crop Factor: 1.5</li><li>SAM Threshold: 0.80</li><li>SAM Mask Hint Threshold: 0.70</li></ul>FreeU_V2:Preset 1: Balance for anime. It makes the picture juicy, and the contour lines are thin and clear.<ul><li>b1 : 1.10</li><li>b2 : 1.10</li><li>s1 : 0.90</li><li>s2 : 0.20</li></ul>Preset 2: For high detail<ul><li>b1 : 1.15</li><li>b2 : 1.20</li><li>s1 : 0.60</li><li>s2 : 0.40</li></ul>Preset 3: Soft and voluminous style (Semi-realism).<ul><li>b1 : 1.05</li><li>b2 : 1.15</li><li>s1 : 0.90</li><li>s2 : 0.10</li></ul>We are looking at color correction, + on large generations, these settings add or subtract multi-level noise. Experiment with the settings, my settings are not a BENCHMARK.<h3 id="compatibility">Compatibility</h3><ul><li>Fully works in ComfyUI</li><li>Fully works in Automatic1111 / Forge / reForge (NOT VERIFIED)</li></ul><h3 id="strengths">Strengths</h3><ul><li>Consistent, high‑quality anatomy and body structure in both portraits and full‑body shots</li><li>Excellent lighting, depth, and volumetric effects</li><li>Distinctive style</li><li>Handles detailed accessories, hair, clothing, and complex scenes well</li><li>Versatile across character art, fashion, and atmospheric compositions</li></ul><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5f1a16ba-8edd-49c1-8579-c9a2aff7924a/original=true/5f1a16ba-8edd-49c1-8579-c9a2aff7924a.jpeg" /><h3 id="notes">Notes</h3>If you're interested in my ComfyUI workflow, let me know, and I’ll break it down and post it for you. I use a 3-stage generation process + FaceDetailer when needed.

HassakuMiXIllustrious

Improved Eye Detail &amp; Iris Clarity: Enhanced generation of the ocular area across all styles. Eyes now feature better reflections, cleaner iris details, and sharper focus, significantly reducing flat or distorted looks.Enhanced 3D Volume &amp; Dimension: Upgraded rendering engine response to lighting prompts. The model now produces better depth, more accurate subsurface scattering, and volumetric shading, making characters and environments pop with a true 3-dimensional feel.Polished Textures &amp; Micro-details: Fine-tuned the processing of complex surfaces such as hair strands, fur, fabrics, and skin textures for overall crisper output.

Files with SHA256: 89127117652a5ed4b0fdcdef9e59dd4d91ab1cf30eae4c6f99c76d56e14e3c1f