<h3>🎨➡️📸 CROWN 2D Anime-to-Real 1:1 Photorealistic Redraw | MOSS-Image Edit</h3>
[English] 
Trained by NineCrowns (<code>[CROWN]</code>) at native <code>1024×1536</code> on an <code>8×NVIDIA B200</code> cluster (<code>Rank 64 / Alpha 64</code>) using dual-pathway MOSS-Image Edit conditioning (<code>Qwen3.5-9B VL</code> vision tokens + <code>Flux2-VAE</code> reference latents initialized from <code>moss_image_edit_lora_rank_64_bf16</code> across 514 curated <code>2D_ANIME -&gt; 1:1_REAL</code> pairs). 
Converts flat 2D Japanese cel-shaded anime illustrations into natural, photorealistic live-action human portraits while preserving 1:1 pose, hairstyle, facial expression, outfit design, and background composition:
<ol>
 <li>True Real-Human Anatomy &amp; Skin Micro-Texture (<code>crown_anime2real, A2R</code>): Resculpts stylized 2D anime eyes, nose, and chin into authentic East Asian / human facial bone structure with natural skin pores, realistic eyelashes, and individual hair strands—avoiding plastic 2.5D doll faces.</li>
 <li>1:1 Structural, Expression &amp; Outfit Fidelity: Preserves exact lip state (closed-mouth vs open-mouth smile), bangs/hairstyle geometry, fabric folds, accessories, and lighting layout from the 2D reference image.</li>
 <li>Native ComfyUI Split Q/K/V Format: Exported in standard ComfyUI <code>diffusion_model.layers.*.proj.{q,k,v,o,ffn_up_gate,ffn_down}</code> key layout for plug-and-play loading.</li>
</ol>
[中文说明] 
由 NineCrowns (<code>[CROWN]</code>) 在 <code>8×NVIDIA B200</code> 集群上以原生 <code>1024×1536</code> 分辨率、<code>Rank 64 / Alpha 64</code> 满血规格精炼的 MOSS-Image Edit 二次元 1:1 像素级重绘成真人 LoRA（融合 <code>Qwen3.5-9B</code> 视觉塔 Token 与 <code>Flux2-VAE</code> 参考图潜变量双路控制，基于 514 组严审去重 <code>2D 动漫 -&gt; 1:1 真人</code> 配对精调）：
<ol>
 <li>真实人类骨相重塑与毛孔级肤质（<code>crown_anime2real, A2R</code>）：将平涂二次元大眼与纸片五官自然重塑为真实人类眼眶、鼻梁骨相、皮肤微肌理与根根分明的真实发丝，拒绝 2.5D 塑料假娃娃脸。</li>
 <li>1:1 表情、发型、穿搭与构图锁定：严格保持原二次元参考图的嘴部开合状态（抿嘴/微笑/露齿笑）、刘海发辫走向、服饰材质纹样与背景光影构图。</li>
 <li>原生兼容 ComfyUI：已拆分为标准 ComfyUI <code>proj.{q,k,v}</code> 独立键位格式，开箱即用。</li>
</ol>
<hr />
📊 Recommended Settings / 推荐推理参数
<ul>
 <li>Base Model (底模): <code>moss_image_sft_bf16.safetensors</code> + <code>qwen3.5_9b_bf16.safetensors</code> (VLM vision enabled) + <code>flux2-vae.safetensors</code></li>
 <li>Resolution &amp; Sampler (分辨率与采样): <code>1024×1536</code> (or <code>1024×1024</code>), <code>28–30 steps</code>, <code>shift: 5.0</code>, <code>CFG: 3.2–3.5</code></li>
 <li>LoRA Weight (推荐权重): <code>0.90 ~ 1.00</code> (All 8 unseen OOD showcase comparisons generated at <code>0.95</code>)</li>
 <li>Trigger Words (触发词): <code>crown_anime2real, A2R</code></li>
</ul>

CROWN 2D Anime-to-Real 1:1 Photorealistic Redraw | MOSS-Image Edit

[EN] v1.0 1024×1536 Rank 64 MOSS-Image Edit release. Converts unseen 2D cel-shaded anime illustrations into 1:1 photorealistic real human portraits while preserving expression, hairstyle, outfit, and composition. Recommended weight: 0.90–1.00. [中文] v1.0 1024×1536 Rank 64 正式版。将任意全新 2D 动漫插画 1:1 像素级重绘为真实人类写真，精准保留表情、发型、服饰与构图。推荐权重：0.90–1.00。

Files with SHA256: 6d7742d41ade2ea9742595fbafe97cdd90751874a98f15b66c933f0489e67fe7