<h1 id="comfyui-mrweazphotolab">ComfyUI-MrWeazPhotoLab</h1>A high-performance ComfyUI custom node created specifically for authentic photography post processing. darkroom color science, physical optical lens bench, 3D depth-aware effects, and micro-detail neural diffusion retouching.GitHub: <a target="_blank" rel="ugc" href="https://github.com/MrWeazelHead/ComfyUI-MrWeazPhotoLab">https://github.com/MrWeazelHead/ComfyUI-MrWeazPhotoLab</a><h3 id="features:">Features:</h3>Interactive Darkroom &amp; Optics Bench UIThe suite features an integrated, high-resolution darkroom console directly on the node canvas:- **Modular Tabbed Interface**: Organized into 5 focused workspaces:1. <code>Retouch &amp; AI</code>: Auto-Masking &amp; SAM Engine, Neural inpaint (external models / checkpoints), pore-locking Frequency Separation, and Subdermal Skin Scattering (SSS).2. <code>Light &amp; Tone</code>: 3D Depth-Aware Studio Relighting, 3D .cube LUT engine, iconic Film Stocks, manual color grading, and filmic shoulder rolloff.3. <code>Optics Bench</code>: Lens distortion curvature (with 24mm, 35mm, 50mm, 85mm prime presets), Optical Depth of Field (with disc bokeh &amp; F-stop apertures), Atmospheric Rayleigh Haze &amp; Silhouette Light Wrap, Pro-Mist diffusion, anamorphic streak flares, multi-spectral halation, lateral chromatic aberration, and $\cos^4(\theta)$ vignetting.4. <code>Film Grain</code>: Photochemical silver halide grain with sensitometric tone response, zero color cast, and 8mm, 16mm, 35mm, 120, and 4x5 format presets. Realistic film grain.5. <code>Master Looks</code>: 22 one-click professional cinematography and film presets across 5 aesthetic categories.- **Collapsible Section Cards (Minimized by Default)**: All control cards are collapsed by default to save canvas space and eliminate visual clutter. Simply click any card header to expand or collapse it with smooth chevron animations <code>▸</code> / <code>▾</code>). Use the top header <code>⊞ Expand All</code> / <code>⊟ Collapse All</code> button to instantly reveal or tuck away all controls.---Auto-Masking &amp; SAM Prompting EngineThe node features a comprehensive automated segmentation and masking engine supporting Segment Anything (SAM):- **Triple Output Architecture <code>image</code>, <code>depth_map</code>, <code>mask</code>)**:- <code>image</code>: The final photochemically and optically processed image.- <code>depth_map</code>: 3D depth map (from ZoeDepth or external input) for depth-aware workflows.- <code>mask</code>: The active segmentation mask formatted as a standard float32 RGB <code>IMAGE</code> <code>[B, H, W, 3]</code>). Connect directly to a ComfyUI <code>PreviewImage</code> or <code>SaveImage</code> node to visualize exactly what SAM or heuristic auto-masking is targeting!- **Detector-Assisted SAM Prompting Engine <code>sam_prompt</code>)**:- Connects natural language prompting directly to SAM using local, zero-internet detectors <code>models/ultralytics/</code>):- **Face, Head &amp; Portrait**: Uses local <code>face_yolov8m</code> to detect exact face bounding boxes, passing boxes directly to SAM <code>predictor.predict(box=box)</code>) for razor-sharp facial isolation.- **Lips &amp; Eyes**: Detects and isolates anatomical facial sub-regions.- **Hair**: Computes crown and lateral bounding bounds relative to the detected face and segments with SAM.- **Hands**: Uses local <code>hand_yolov8s</code> to detect hands and fingers.- **Clothes &amp; Torso**: Uses local <code>yolov8n</code> to detect person bodies and isolates torso/clothing bounds.- **80 COCO Objects**: Prompts like <code>dog</code>, <code>cat</code>, <code>car</code>, <code>chair</code>, <code>bottle</code>, <code>cup</code>, <code>laptop</code>, <code>phone</code>, etc., locate matching objects via <code>yolov8n</code> and segment them with SAM.- **Background &amp; Sky**: Prompts containing <code>background</code>, <code>backdrop</code>, <code>sky</code>, <code>wall</code>, or <code>room</code> detect foreground subjects and automatically invert the mask.- **Coordinates**: Enter normalized coordinate pairs (e.g. <code>0.35, 0.45</code>) or use the UI center sliders.- **Auto-Activation**: Typing any prompt in <code>sam_prompt</code> automatically activates SAM without needing to flip the mode dropdown from <code>Disabled</code>.- **Quick Target Chips in UI**: One-click buttons in the darkroom UI for <code>[👤 Subject, 😊 Face, 💇 Hair, 👄 Lips, ✋ Hands, 👕 Clothes, 🐕 Pet, 🚗 Car, 🌄 Background]</code>.- **Depth-Assisted Center Snapping <code>sam_depth_assist</code>)**: Uses local ZoeDepth to automatically locate the nearest foreground subject around the prompt coordinates—no need to manually pinpoint pixel coordinates.- **Algorithmic Auto-Mask Modes**:- <code>Depth Foreground Isolation</code>: Isolates subjects using metric depth.- <code>Skin &amp; Portrait Tones</code>: Segments skin tones and facial features for retouching.- <code>Specular Highlights</code>: Isolates high-luma glints and reflections.- <code>Shadows &amp; Blacks</code>: Targets low-luma shadow regions.- **Configurable Mask Scope**:- <code>Retouch Only</code>: Restricts Diffusion Inpaint and Frequency Separation to the mask, while optical/darkroom effects apply globally.- <code>All Effects</code>: Completely isolates **all** enabled effects (retouching, relighting, LUTs, grain, mist, flares, etc.) to the masked area. Non-masked pixels remain 100% bit-for-bit identical to the input image.- **Empty Mask Bypass**: If a mask is provided or generated with all zeros, all processing and effects are automatically bypassed, immediately returning the untouched input image and a zero mask.- **Mask Inversion &amp; Feathering**: Instant 1-click mask inversion and smooth edge feathering with Gaussian blur.---Automatic 3D Depth Mapping Engine (ZoeDepth)The node integrates local neural depth estimation powered by <code>Intel/zoedepth-nyu-kitti</code>:- **100% Local &amp; Private**: Runs entirely on your local machine using cached HuggingFace weights without any internet requests.- **Plug-and-Play Zero Setup**: If no <code>depth_map</code> socket is connected, the node automatically estimates high-fidelity metric 3D depth for your image. You can also wire external depth maps (e.g., Depth Anything V2) directly into the <code>depth_map</code> socket.- **Multi-Section Depth Integration**:- **Depth of Field (DoF)**: Real circular disc convolution bokeh with physical aperture presets <code>f/1.2</code> through <code>f/22</code>) and auto-focus tracking on foreground subjects.- **3D Studio Relighting**: True 3D spherical inverse-square falloff $(X, Y, Z)$ where the light source can be positioned in front of or behind subjects for realistic key, fill, or rim lighting.- **Atmospheric Aerial Haze**: Realistic Rayleigh distance scattering that smoothly fades distant mountains and horizons while keeping foreground subjects crisp.- **Silhouette Light Wrap**: Edge-detected light bleeding around foreground silhouettes, hair strands, and shoulders.- **Dedicated Outputs**: Returns both <code>image</code>, <code>depth_map</code>, and <code>mask</code> for modular compositing.<h3 id="installation-and-setup">Installation &amp; Setup</h3>1. Place this repository into your ComfyUI <code>custom_nodes</code> directory:```bashcustom_nodes/ComfyUI-MrWeazPhotoPro/```2. Place any <code>.cube</code> LUT files into <code>models/luts/</code>3. In ComfyUI, add node: *<code>MrWeaz - PhotoLab</code>** (under category <code>MrWeazNodes/PhotoLab</code>).

MrWeaz - PhotoLab Node

-- Added AWB/Dynamic Range algorithms for better colour correction. -- Added more presets and film stock profiles.GitHub: <a target="_blank" rel="ugc" href="https://github.com/MrWeazelHead/ComfyUI-MrWeazPhotoLab">https://github.com/MrWeazelHead/ComfyUI-MrWeazPhotoLab</a>

Files with SHA256: a33b08f9fa6c841ff62ac039c840ff77904267264c82f302d043e6c147e34c7a