<h3 id="*showcase-images-include-embedded-workflow-and-are-post-processed-with-sharpening-and-noise-grain.">*Showcase images include embedded workflow and are post-processed with sharpening and noise grain.</h3>In other words, your results may vary depending on your workflow.<hr /><h3 id="recommended-settings-are-below.">Recommended settings are below.</h3><hr /><h1 id="version-features">Version features</h1><h3 id="rather-than-improvements-each-version-has-their-own-strengths">Rather than improvements, each version has their own strengths</h3><ul><li>v1.1 - <s>Most realistic</s> 2nd most realistic</li><li>v1.2 - Semi realistic - sometimes include outline</li><li>v1.33 - Slightly less realistic than 1.1 in exchange for better lighting/composition/prompt adherence.</li><li>v1.34 - v1.33 + slightly more detail in background</li><li>v2.4 - Most realistic</li><li>v3.1 - Most balanced</li></ul><h3 id="recommendation">Recommendation</h3><ul><li>From realistic to semi realistic = v2.4 &gt; v1.1 &gt; v3.1 &gt; v1.34/1.33 &gt; 1.2</li><li>Personal favorite = v3.1 / v2.4 </li></ul><hr /><h1 id="version-notes">Version notes</h1><h3 id="v3.1">v3.1</h3><ul><li>Most balanced version - combining strengths from previous versions</li><li>Realistic skin texture, good contrast/lighting/colour, better background details</li><li>Good face proportions</li><li>Good prompt adherence</li><li>Aesthetic more illustrative (Closer to v1.34 than v2.4)</li></ul><h3 id="v2.4">v2.4</h3><ul><li>Focus more on realism</li><li>More realistic skin texture</li><li>Should have better prompt adherence</li><li>Aesthetic closer towards realistic photography than illustrative (Compared to v.1.0 variants)</li></ul><h3 id="v1.34">v1.34</h3><ul><li>Should have slightly better details to the background.</li></ul><h3 id="v1.33">v1.33</h3><ul><li>Focus on keeping realism while improving composition, background and lighting compared to v1.1.</li><li>Less creative/aesthetic than 1.2 in terms of background.</li></ul><h3 id="v1.2">v1.2</h3><ul><li>More focus towards semi realism *includes outline at certain times</li><li>Better lighting, composition, background, aesthetic</li><li>Structure feels more stable compared to v1.1</li></ul><hr /><h1 id="model-focus">Model focus</h1><ul><li>3DCG features with more detailed skin texture. (Heavily dependent on heun sampler, not recommended if using web generator or if you don't know what you're doing)</li><li>Face proportions closer to real life. Less anime like.</li></ul><h3 id="strengthquirks">Strength/Quirks</h3><ul><li>Facial features can respond to prompts</li></ul><h3 id="weakness">Weakness</h3><ul><li>Not suitable for male because of female face modifier.</li><li>Will do beauty mark/mole under eye (Apparent in lower cfg; use negative prompt)</li><li>Skin pores can look too smooth/uniform/unnatural texture depending on sampler/scheduler.</li><li><s>Backgrounds (res_multistep + normal with high steps 50-100 - has chance to resolve scale issues but might introduce artifacts) </s>(better in version 1.2 and above)</li></ul><hr /><h1 id="recommended-settings">Recommended settings</h1><h3 id="(can-see-workflow-via-uploaded-images)">(can see workflow via uploaded images)</h3><h2></h2><h3 id="positive-prompts">positive prompts</h3>can change the face structure/matureness<pre><code>milf, adult</code></pre>makeup/beauty enhancements<pre><code>(dark maroon eyeshadow under eye, maroon blush makeup, eyeliner:1), (doe eyes makeup:0.5), shiny skin, (perfect hands:0.9)</code></pre>realism tags<pre><code>subsurface scattering, photorealistic, realistic, hyperrealistic</code></pre><h3 id="initial-sampler">initial sampler</h3>sa_solver_pece + sgm_uniform<ul><li>sweet spot: steps 30-40, cfg 6-7 @ 1024x1280 pixels</li><li>more realistic at the cost of contrast/dark/warmer image: steps 30-40, cfg 4-5</li><li>provides the most stable structure, details and subtle noise at reasonable speed. handles cfg burn not bad as well. use cfg higher than 7 for more contrast.</li></ul>dpm_adaptive + exponential<ul><li>steps doesn't matter since adaptive, cfg 4-10</li><li>slower but very reliable without much errors. however compared to sa_solver_pece, it is smoother/no noise and lack small details.</li></ul>dpmpp_sde + sgm_uniform/exponential/sgm_uniform<ul><li>steps 30, cfg 4-7</li><li>feels slightly faster than the two above with reasonable structure.</li></ul>res_multistep + sgm_uniform/normal/simple<ul><li>steps 30-100, cfg 4-6</li><li>fast, reliable but slightly boring/simplistic results</li><li>res_multistep is fast and converges at high steps, fixes the structural problems 90% of the time.</li></ul><h3 id="upscalehiresfix">upscale/hiresfix</h3>upscale model<ul><li>any</li><li>realesrgan_x2 (only because it's fast - don't think it makes a big difference which ever used)</li></ul>heun + normal/ddim_uniform(second choice)<ul><li>steps 15-20, cfg 3-4, denoise 0.35-0.55</li><li>more random noise pattern with cleaner details compared to 2m_sde_heun karras</li></ul>dpmpp_2m_sde_heun + karras<ul><li>steps 20-30, cfg 3-5, denoise 0.35-0.55</li><li>slightly faster than above, noise pattern more uniform compared to heun</li></ul><h3 id="facedetailer">facedetailer</h3>heun + normal<ul><li>steps 10, cfg 2.5-6, denoise 0.25-0.4</li></ul>dpmpp_2m_sde_heun + karras<ul><li>steps 10-15, cfg 2.5-6, denoise 0.25-0.4</li><li>mole under eye might appear if low cfg</li></ul><hr /><h1 id="remarks">Remarks</h1><ul><li>sa_solver_pece + sgm_uniform is my new favourite combination</li><li>usual euler/2m/3m + sgm uniform/normal/karass should work though</li><li>i prefer heun as it produces natural noise/texture at lower step values and loses it at higher (recommended schedulers: normal/ddim_uniform/ays+)</li><li>for prompt adherence: higher resolution can have lower cfg, lower resolution need higher cfg</li><li>personal frequently used resolution before upscale: <s>896x1120</s> 1024x1280 or 1152×1440 (more pixels = more space for details like hands)</li><li>post-processing nodes with sharpening, bloom, grain, tone curves for more texture in final image. (see workflow in images)</li></ul>

FUBUWA MIX 3D REAL

fubuwaMIX3DREAL_v11.safetensors

v1.1

Files with SHA256: 0656b4fe669b22c3a27202afafc75c3c5bc3f7986312cec15759cbce5134d58a