[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-81106":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":14,"stars7d":16,"stars30d":17,"stars90d":15,"forks30d":15,"starsTrendScore":18,"compositeScore":19,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":9,"pushedAt":9,"updatedAt":24,"readmeContent":25,"aiSummary":26,"trendingCount":15,"starSnapshotCount":15,"syncStatus":27,"lastSyncTime":28,"discoverSource":29},81106,"ComfyUI-BSS_FLSampler","BlackSnowSkill\u002FComfyUI-BSS_FLSampler","BlackSnowSkill","An advanced Foveated Latent Sampler (FLS) for ComfyUI. Selective high-frequency local contrast boosting and stochastic texture noise injection based on latent dynamics. Features a gorgeous premium dark matte gold UI by BSS.",null,"JavaScript",64,3,38,1,0,10,26,4,48.41,false,"master",true,[],"2026-06-12 04:01:32","# FLSampler (Foveated Latent Sampler) (BSS)\n\n🇷🇺 [Читать на русском языке](README_RU.md)\n\n**FLSampler (BSS)** is a high-performance custom sampler for ComfyUI, inspired by the physiology of human sight (**Foveated Vision**). It is engineered to deliver stunning cinematic clarity, rich micro-textures, and precise depth-of-detail control directly within the latent space, without adding any overhead to generation times.\n\n> [!IMPORTANT]\n> **Author and Developer:** **blacksnowskill (BSS)**  \n> **© 2026 blacksnowskill (BSS). All rights reserved.**  \n> This project is protected by copyright. Any unauthorized copying, modification without attribution, or representing this code as your own product is strictly prohibited.\n\n> [!TIP]\n> **Ultimate Speed & Performance Booster:**  \n> To achieve the ultimate balance of breathtaking visual clarity and lightning-fast generations, we highly recommend pairing **FLSampler (BSS)** with our core optimization suite **[ANIMA_BOOSTER](https:\u002F\u002Fgithub.com\u002FBlackSnowSkill\u002FANIMA_BOOSTER)**. It delivers a massive **3.5× to 5.0× speedup** on Anima DiT models, while FLSampler perfectly preserves and enhances every tiny high-frequency detail!\n\n---\n\n## 🔬 Physics & Mathematics of the Process\n\nStandard samplers (KSampler) apply noise injection and denoising uniformly across the entire latent space, which often results in either blurred micro-textures (plastic look) or over-sharpened artifacts (burnt images).\n\n**FLSampler** resolves this bottleneck through intelligent, selective latent modification:\n\n1. **Differential Latent Analysis (Focus Detection):**\n   At each denoising step, the sampler computes the absolute difference $\\Delta = |x_{0, t} - x_{0, t-1}|$ between the current predicted clean image $x_0$ and its state from the previous step. This mathematically pinpoints the exact regions where geometry, object boundaries, and fine structures are currently forming.\n2. **Adaptive Fovea Map with Inertia (Momentum Mask):**\n   The raw difference map is smoothed using a 2D average pooling operator (`avg_pool2d`), after which a dynamic threshold is calculated based on its mean and standard deviation:\n   $$Threshold = \\mu + 0.5 \\cdot \\sigma$$\n   A soft, organic focus mask (**Fovea Mask**) is then generated using a sigmoid activation. To prevent erratic focal jumps between steps, a momentum coefficient (**Momentum** \u002F `mask_inertia`) is applied, ensuring a smooth, natural transition of detail zones.\n3. **Local Contrast (Sharpness Boost):**\n   Inside the high-activity Fovea zone, an **Unsharp Masking** operation is performed directly on the latent tensor. High frequencies are extracted by subtracting a blurred version of the latent from the original, and then selectively amplified. This produces incredibly crisp object boundaries and realistic cinematic depth of field.\n4. **Stochastic Texture Injection (Micro-Grain):**\n   During the early-to-mid stages of generation, a controlled micro-noise tensor is injected solely into the active Fovea zones. The subsequent denoising steps naturally resolve this noise into ultra-fine high-frequency textures (skin pores, clothing fabrics, wood grains, foliage, hair, or fur), preventing flat plastic surfaces. The injection smoothly decays towards the final steps, guaranteeing a clean, high-fidelity output.\n\n---\n\n## 🎨 BSS Premium UI\n\nThe node comes integrated with the luxurious **BSS Premium UI** design system, fully synchronized with the **ANIMA_BOOSTER** aesthetic:\n* **Luxury Matte Theme:** A deep, matte ultra-dark graphite body (`#0f0f0eff`) that seamlessly blends inputs, outputs, and the node itself.\n* **Rich Gold Accents:** Harmonious gold sliders (`#d4af37`), gold knobs (`#e5c158`), gold dropdown arrowheads, and gold-trimmed numeric input borders.\n* **Complete Silence:** Absolute suppression of annoying hover tooltips during widget adjustment and node dragging.\n* **Branding:** A fine gold divider bar situated above the first widget and a sharp, metallic `\"BSS OPTIMIZED\"` branding engraving in the bottom-right corner.\n\n---\n\n## 🎛️ Parameters & Impact Guide\n\nThis node completely replaces the standard KSampler, introducing both classic controls and proprietary physical regulators:\n\n| Parameter | Range | Recommended | Description |\n| :--- | :--- | :--- | :--- |\n| **fovea_strength** | `0.0 – 10.0` | **`3.0 – 8.0`** | **Texture Strength:** Controls high-frequency detail, skin porosity, and fine textures. Set to `0.0` to disable noise injection entirely. Default: `3.0`. |\n| **sharpness** | `0.0 – 3.0` | **`0.3 – 1.0`** | **Local Contrast:** Dictates edge definition and object boundary clarity in focus zones. Delivers a \"tack-sharp\" look without halo artifacts or \"frying\" the image. Default: `0.5`. |\n| **mask_inertia** | `0.0 – 0.99` | **`0.80 – 0.90`** | **Mask Inertia:** Focus movement smoothness between steps. Values `>0.85` stabilize detail zones, while lower values force the focus to react instantly to sudden changes. Default: `0.85` (step `0.01`). |\n| **denoise** | `0.0 – 1.0` | **`1.0`** | Standard denoising strength. |\n| **steps** \u002F **cfg** | - | - | Standard generation steps and prompt guidance scale. |\n\n---\n\n## 📊 Node Outputs\n\nThe node provides two outputs:\n1. 🖼️ **latent**: The optimized latent tensor ready for VAE decoding.\n2. 👁️ **fovea_mask**: The accumulated Fovea focus map visualized as an RGB image. You can pipe this directly into a Preview Image or Save Image node to visually inspect exactly where the sampler focused its computation during the run—an unprecedented debugging tool for complex workflows!\n\n---\n\n## 📋 Recommended Workflow\n\n```\n[ MODEL (e.g., Anima \u002F SDXL) ] ──> [ positive ]\n                               ──> [ negative ]\n                                        ↓\n[ Latent Image (resolution) ] ─────> [ latent_image ] ──> [ FLSampler (BSS) ] ──(latent)──> [ VAE Decode ] ──> [ Save Image ]\n                                                                 └──(fovea_mask)──> [ Preview Image ]\n```\n\n---\n\n## 📥 Installation\n\n1. Open your terminal in the `ComfyUI\u002Fcustom_nodes\u002F` directory.\n2. Clone the repository:\n   ```bash\n   git clone https:\u002F\u002Fgithub.com\u002FBlackSnowSkill\u002FComfyUI-BSS_FLSampler.git\n   ```\n3. Restart ComfyUI and refresh your browser cache (`Ctrl + F5`) to enjoy the gorgeous BSS Premium UI.\n\n---\n\n## ☕ Support & Development\n\nIf you love my work and want to support the development of future optimizations, nodes, and custom models, please consider supporting me:\n\n- **Boosty**: [Support & Exclusive Models](https:\u002F\u002Fboosty.to\u002Fblacksnowskill)\n\n---\n\n## 📄 License & Usage\n\n© 2026 blacksnowskill (BSS). All rights reserved.\n\nThis software is an experimental release. Feedback is highly welcome.\n**Notice:** This project is protected by copyright. Any unauthorized copying, distribution, merging with other projects, or hosting on other repositories\u002Fwebsites without the explicit written permission of the author is strictly prohibited.\n","FLSampler (Foveated Latent Sampler) 是一个为 ComfyUI 设计的高性能自定义采样器，灵感来源于人类视觉的中心凹视觉机制。该项目通过选择性地增强高频局部对比度和基于潜在动态的随机纹理噪声注入，实现了出色的电影级清晰度、丰富的微纹理以及精确的细节深度控制，且不会增加生成时间的开销。技术上，FLSampler 采用差分潜变量分析、自适应焦点图与惯性处理、局部对比度增强及随机纹理注入等方法，确保图像在关键区域保持高清晰度的同时避免过度锐化或模糊。适用于需要高质量图像生成但又不希望牺牲速度的场景，如创意设计、数字艺术创作等。",2,"2026-06-11 04:03:33","CREATED_QUERY"]