Tired of constantly doing mental math, checking resolution charts, or getting distorted images because your dimensions don't match your model's native training?The Aspect Ratio Master node provides a seamless, intelligent, and highly precise resolution management utility directly inside your ComfyUI interface. Built for creators who want perfect compositions without the headache, it eliminates guesswork by embedding native model resolutions and automated quality tiers directly into your workflow.<h2 id="key-features">🚀 Key Features</h2><ul><li>All Major Models Pre-Configured: Out-of-the-box support for the most popular AI models (including SD 1.5, SDXL, Flux, and more). It automatically knows their optimal native base resolutions, preventing unwanted stretching, bad framing, or generation artifacts.</li><li>Next-Gen Support (2026): Full native configuration for NVIDIA Cosmos (Gen/Predict), SANA-1.5 / SANA-Video, Wan Video (V2.x/V2.5), Hunyuan Video, CogVideoX, Mochi 1, HiDream, LTX Video / LTX-2, Z-Image Family, Flux.1/2, Krea 2, AuraFlow, Lumina, OmniGen / OmniGen2, SD 3.x, Hunyuan Image 1.0 / 2.1, SDXL, Pony, Illustrious / NoobAI, SD 1.5 — each with its correct latent channel count and VAE compression factor.</li><li>Correct Latent Channels per Architecture: Each model outputs the exact right latent shape — <code>4ch</code> for SDXL/SD1.5, <code>12ch</code> for Mochi, <code>16ch</code> for Flux/Wan/SD3/CogVideoX/etc., <code>32ch</code> for SANA, <code>128ch</code> for LTX Video — preventing channel mismatch crashes with the VAE decoder.</li><li>Dynamic Quality Modes (Low, Middle, High): Switch effortlessly between Low, Middle, and High quality presets. The node automatically scales and recalculates your resolution presets on the fly to match your chosen target fidelity.</li><li>Smart Ratio Calculations: Quickly toggle between standard aspect ratio presets (16:9, 9:16, 1:1, 4:3, 2:3, etc.) or define your own custom aspect ratios instantly.</li><li>Native &amp; Fluid UI Integration: Powered by a dedicated front-end script (<code>aspect_ratio_master.js</code>), making preset swapping and quality shifting interactive and immediate within your graph.</li><li>Zero Performance Footprint: Clean backend logic (<code>aspect_ratio_master.py</code>) ensuring rapid execution with absolutely no VRAM overhead.</li></ul>📦 Note: This node is included as part of the ComfyUI-3R3BOS-Pack suite. Installing the pack gives you instant access to this layout tool along with our other workflow optimization utilities.<h2 id="installation">🛠️ Installation</h2><h3 id="method-1:-via-comfyui-manager-(recommended)">Method 1: Via ComfyUI Manager (Recommended)</h3><ol><li>Open the ComfyUI Manager.</li><li>Click on Custom Nodes Manager.</li><li>Search for <code>ComfyUI-3R3BOS-Pack</code>.</li><li>Click Install, then restart your ComfyUI instance.</li></ol><h3 id="method-2:-manual-installation-(git)">Method 2: Manual Installation (Git)</h3>Navigate to your <code>custom_nodes</code> directory and clone the repository:Bash<pre><code>cd ComfyUI/custom_nodes
git clone https://github.com/3R3BOS/ComfyUI-3R3BOS-Pack.git
</code></pre>Restart ComfyUI to let the interface register the custom scripts.<h2 id="useful-links-and-resources">🔗 Useful Links &amp; Resources</h2>✨ Star the repository to stay updated with the latest features and optimizations!<ul><li>📁 <a target="_blank" rel="ugc" href="https://github.com/3R3BOS/ComfyUI-3R3BOS-Pack">GitHub Repository — ComfyUI-3R3BOS-Pack</a></li><li>🚀 <a target="_blank" rel="ugc" href="https://civarchive.com/models/2298795/comfyui-3r3bos-pack">Main Pack Page on Civitai</a></li></ul><h2 id="quick-start-guide">💡 Quick Start Guide</h2><ol><li>Right-click on the workspace -&gt; <code>Add Node</code> -&gt; <code>3R3BOS</code> -&gt; <code>Aspect Ratio Master</code>.</li><li>Select your Model: Pick the architecture you are currently using (e.g., SDXL, Flux) to lock in native base resolutions.</li><li>Set your Quality Tier: Choose <code>Low</code> for fast drafting, <code>Middle</code> for standard generation, or <code>High</code> for maximum fidelity. The presets adapt automatically!</li><li>Choose your Aspect Ratio: Select your desired preset or enter a custom ratio.</li><li>Connect the <code>width</code> and <code>height or latent</code> output slots directly into your <code>Empty Latent Image</code> node (or your generation node of choice) and queue your prompt!</li></ol>

Aspect Ratio Master 📐 (ComfyUI-3R3BOS-Pack)

aspectRatioMaster_v100.zip

v1.0.0

Tired of frantically switching between tabs or awkwardly overlaying images just to spot the differences in your renders?The Image Comparer Slider node delivers a smooth, interactive, and ultra-precise comparison slider directly inside your ComfyUI interface. Built for power users who demand absolute control, it lets you inspect upscale quality, evaluate different checkpoints, or scrutinize fine details without ever breaking your workflow.<h2 id="key-features">🚀 Key Features</h2><ul><li>Real-Time Interactive Comparison: Smoothly slide the divider left and right for a seamless, transparent transition between both renders.</li><li>Synchronized Zoom &amp; Pan (Absolute Pixel-Peeping): Scroll your mouse wheel to zoom and drag to pan. Both images remain perfectly locked and synchronized, allowing you to track minute variations in grain, sharpness, or artifacts at the exact same pixel coordinates.</li><li>Native &amp; Fluid UI Integration: Powered by a dedicated front-end implementation (<code>image_comparer_slider.js</code>), the canvas wrapper integrates flawlessly into the ComfyUI graph with zero latency.</li><li>VRAM Lightweight: Engineered with clean backend logic (<code>image_comparer_slider.py</code>) to ensure it leaves no performance footprint on your heavy generation pipelines.</li></ul>📦 Note: This node is included as part of the ComfyUI-3R3BOS-Pack suite. Installing the pack gives you instant access to this comparison tool along with our other workflow optimization utilities.<h2 id="installation">🛠️ Installation</h2><h3 id="method-1:-via-comfyui-manager-(recommended)">Method 1: Via ComfyUI Manager (Recommended)</h3><ol><li>Open the ComfyUI Manager.</li><li>Click on Custom Nodes Manager.</li><li>Search for <code>ComfyUI-3R3BOS-Pack</code>.</li><li>Click Install, then restart your ComfyUI instance.</li></ol><h3 id="method-2:-manual-installation-(git)">Method 2: Manual Installation (Git)</h3>Navigate to your <code>custom_nodes</code> directory and clone the repository:Bash<pre><code>cd ComfyUI/custom_nodes
git clone https://github.com/3R3BOS/ComfyUI-3R3BOS-Pack.git
</code></pre>Restart ComfyUI to let the interface register the custom scripts.<h2 id="useful-links-and-resources">🔗 Useful Links &amp; Resources</h2>✨ Star the repository to stay updated with the latest features and optimizations!<ul><li>📁 <a target="_blank" rel="ugc" href="https://github.com/3R3BOS/ComfyUI-3R3BOS-Pack">GitHub Repository — ComfyUI-3R3BOS-Pack</a></li><li>🚀 <a target="_blank" rel="ugc" href="https://civarchive.com/models/2298795/comfyui-3r3bos-pack">Main Pack Page on Civitai</a></li></ul><h2 id="quick-start-guide">💡 Quick Start Guide</h2><ol><li>Right-click on the workspace -&gt; <code>Add Node</code> -&gt; <code>3R3BOS</code> -&gt; <code>Image Comparer Slider</code>.</li><li>Connect your baseline image to the <code>image_1</code> input slot.</li><li>Connect your modified/upscaled image to the <code>image_2</code> input slot.</li><li>Queue the prompt.</li><li>Inspect away: Use your mouse wheel to zoom, left-click and drag to move around, and drag the center bar to compare!</li></ol>

Image Comparer Slider ↔️ (ComfyUI-3R3BOS-Pack)

imageComparerSlider_v100.zip

<h1 id="comfyui-3r3bos-pack">ComfyUI 3R3BOS Pack</h1>The ultimate workflow engineering toolkit designed to bring professional control, "Human-in-the-Loop" efficiency, and zero-latency visualization to your ComfyUI environment.Stop wasting GPU cycles on bad batches, doing mental math for model resolutions, or losing your favorite prompts in external Notepad files. The ComfyUI 3R3BOS Pack unifies your workflow with a sober, monochrome, native canvas aesthetic built for power users.<h2 id="the-toolkit">🚀 The Toolkit</h2><h3 id="1.-prompt-selector-your-interactive-prompt-brain">🧠 1. Prompt Selector — Your Interactive Prompt Brain</h3>Stop copying and pasting. This node turns your canvas into a fully interactive local prompt database, making text management dynamic and completely modular.<ul><li>In-Canvas Customization: Built-in visual editor lets you create, rename, or delete categories directly inside the ComfyUI interface.</li><li>Organized Tagging: Easily store complex prompts, aesthetic styles, or LoRA trigger words with a custom color-coded system (e.g., tags for NSFW, custom themes, or styles).</li><li>Live Workspace Control: Features a real-time prompt preview panel and a live override engine, letting you tweak text on the fly without breaking your core database structure.</li></ul><h3 id="2.-batch-selector-the-ultimate-visual-gatekeeper">🛑 2. Batch Selector — The Ultimate Visual Gatekeeper</h3>Filter your generations like a pro before committing resources. Replaces complex preview-and-cancel loops with hard algorithmic control over your execution pipeline.<ul><li>True Native UI Grid: Draws a responsive image grid directly onto the node canvas with Zero HTML overlays—completely eliminating browser layout issues and scaling mismatches.</li><li>Resource Optimization: Pauses workflow execution dynamically, letting you cherry-pick only the successful generations to pass on to expensive upscalers, refiners, or face-swappers.</li><li>Bulletproof Core Stability: Engineered with a built-in "Heartbeat" mechanism and crash-protection logic to prevent workflow freezing when switching browser tabs or handling corrupted image packets.</li><li>Instant Hard Cancel: Features a dedicated, instant-kill canvas button to immediately halt the workflow if the base batch fails to meet expectations.</li></ul><h3 id="3.-image-comparer-slider-absolute-pixel-peeping">⚡ 3. Image Comparer Slider — Absolute Pixel-Peeping</h3>The gold standard for real-time A/B testing. Compare checkpoints, test subtle LoRA weights, or examine before/after upscales with flawless clarity.<ul><li>Synchronized Zoom &amp; Pan: Scroll your mouse wheel to zoom and drag to pan across your composition. Both sides remain perfectly locked and synchronized, tracking grain, sharpness, and artifacts at identical pixel coordinates.</li><li>Zero-Lag Scrubbing: Powered by specialized client-side caching to maintain a locked, fluid 60fps performance during side-by-side divider scraping.</li><li>Dynamic Auto-Compaction: Hook up to 20 individual images simultaneously; the node dynamically opens new input slots and auto-compacts itself to keep your graph meticulously clean.</li></ul><h3 id="4.-aspect-ratio-master-the-definitive-resolution-engine">🎯 4. Aspect Ratio Master — The Definitive Resolution Engine</h3>Say goodbye to stretched proportions and generation artifacts. This tool completely automates resolution math for cutting-edge generative architectures.<ul><li>Modern Model Database: Out-of-the-box native resolution and configuration blueprints for modern models like Flux.2, Hunyuan 2.1 (2K), Wan 2.2, and LTX Video.</li><li>Next-Gen Support (2026): Full native configuration for NVIDIA Cosmos (Gen/Predict), SANA-1.5 / SANA-Video, Wan Video (V2.x/V2.5), Hunyuan Video, CogVideoX, Mochi 1, HiDream, LTX Video / LTX-2, Z-Image Family, Flux.1/2, Krea 2, AuraFlow, Lumina, OmniGen / OmniGen2, SD 3.x, Hunyuan Image 1.0 / 2.1, SDXL, Pony, Illustrious / NoobAI, SD 1.5 — each with its correct latent channel count and VAE compression factor.</li><li>Correct Latent Channels per Architecture: Each model outputs the exact right latent shape — <code>4ch</code> for SDXL/SD1.5, <code>12ch</code> for Mochi, <code>16ch</code> for Flux/Wan/SD3/CogVideoX/etc., <code>32ch</code> for SANA, <code>128ch</code> for LTX Video — preventing channel mismatch crashes with the VAE decoder.</li><li>Dynamic Quality Tiers: Features an automated preset system toggle (Low, Middle, High Quality) that scales and recalculates target resolutions instantly depending on your current pipeline phase (drafting vs. final rendering).</li><li>Labs Mode 🧪: Unlocks optimized community-tested "Sweet Spot" configurations (such as the premium 1.5MP Flux tuning) for extreme output fidelity.</li><li>Visual Aspect Grid: Renders a beautiful, real-time graphical preview of your bounding box directly on the node surface.</li></ul><h2 id="important-data-safety">⚠️ Important — Data Safety</h2>Your Prompt Selector library is stored entirely locally on your machine. To ensure you never lose your data, pay close attention to the directory file below:📂 Database Location: <code>ComfyUI/custom_nodes/ComfyUI-3R3BOS-Pack/prompts_db.json</code>👉 Critical Action: Always manually back up this file before updating the pack, performing clean reinstalls of ComfyUI, or executing a fresh Git pull from GitHub. Losing this JSON file will permanently clear your custom prompt library.<h2 id="installation">🛠️ Installation</h2><h3 id="method-1:-via-comfyui-manager-(recommended)">Method 1: Via ComfyUI Manager (Recommended)</h3><ol><li>Open the ComfyUI Manager menu.</li><li>Click on Custom Nodes Manager.</li><li>Search for <code>3R3BOS Pack</code>.</li><li>Click Install, and restart your ComfyUI server.</li></ol><h3 id="method-2:-manual-installation-(git-clone)">Method 2: Manual Installation (Git Clone)</h3>Open your terminal inside your local custom nodes directory and execute:Bash<pre><code>cd ComfyUI/custom_nodes
git clone https://github.com/3R3BOS/ComfyUI-3R3BOS-Pack.git
</code></pre>Restart ComfyUI to allow the frontend scripts to cache into your environment.<h3 id="method-3:-direct-manual-archive-(.rar)">Method 3: Direct Manual Archive (.rar)</h3><ol><li>Download the <code>.rar</code> package directly from this page.</li><li>Extract the archive contents.</li><li>Move the extracted folder named <code>ComfyUI-3R3BOS-Pack</code> directly into your <code>ComfyUI/custom_nodes/</code> directory.</li><li>Restart your ComfyUI server instance.</li></ol><h2 id="links-and-support">🔗 Links &amp; Support</h2>⭐ Please leave a star rating on GitHub if you enjoy using the suite—it significantly helps support continued development! Thank you! ❤️<ul><li>📁 <a target="_blank" rel="ugc" href="https://github.com/3R3BOS/ComfyUI-3R3BOS-Pack">Official GitHub Repository</a></li><li>🐛 <a target="_blank" rel="ugc" href="https://github.com/3R3BOS/ComfyUI-3R3BOS-Pack/issues">Submit Bug Reports &amp; Feature Requests</a></li></ul><h2 id="full-update-log">📋 Full Update Log</h2><h3 id="v1.0.8">v1.0.8</h3><ul><li>CRITICAL FIX — Aspect Ratio Master: Latent channel count was hardcoded to <code>4</code> for every architecture. All modern models (Flux, Wan, SD3, CogVideoX, Hunyuan Video, Cosmos, etc.) were generating an incorrect latent shape, causing VAE decode crashes or silent grey outputs. Each architecture now outputs the exact correct channel count: <code>4ch</code> (SDXL/SD1.5), <code>12ch</code> (Mochi 1), <code>16ch</code> (Flux/Wan/SD3/CogVideoX/HiDream/Krea2/AuraFlow/Lumina/Z-Image/OmniGen/Hunyuan Image/Cosmos), <code>32ch</code> (SANA-1.5), <code>128ch</code> (LTX Video).</li><li>FIX — VAE Compression Factor: <code>LTX Video</code> corrected from <code>vae_factor=8</code> to <code>vae_factor=32</code> (DC-VAE). <code>Hunyuan Image 2.1</code> corrected from <code>vae_factor=8</code> to <code>vae_factor=32</code> (32× spatial compression). <code>PixArt Alpha/Sigma</code> corrected to <code>4ch</code> (uses SDXL-compatible VAE, not a 16ch architecture).</li></ul><h3 id="v1.0.7">v1.0.7</h3><ul><li>PREMIUM SLATE-CYAN THEME: Complete visual redesign of both Aspect Ratio Master and Image Comparer Slider with a unified premium Slate-Steel grey theme (<code>#242730</code>) and soft cyan blue accents (<code>#00b4d8</code>).</li><li>ASPECT RATIO MASTER REFACTOR: Integrated support for modern 2026 architectures (NVIDIA Cosmos, SANA-1.5, Wan Video, Z-Image, OmniGen), direct custom pixel inputs, real-time VAE latent shape tracking, canvas shadow bleed isolation, and layout dimension locks.</li><li>IMAGE COMPARER SLIDER OVERHAUL: Added a Synchronized Texture Inspector (Pointer Pan &amp; Zoom), Auto-Play Ping-Pong loop functionality, dynamic source node name tracking, memory-safe pointer capture handlers, and automated cache cleanup routines.</li></ul><h3 id="v1.0.6">v1.0.6</h3><ul><li>NEW NODE: <code>Prompt Selector</code> launched as a major workflow expansion feature.</li><li>FEATURES: Implemented a comprehensive prompt database manager with real-time text editing, color-coded custom categories, and premium canvas-native interface modules.</li></ul><h3 id="v1.0.5">v1.0.5</h3><ul><li>CRITICAL FIX: Resolved a breaking issue where the <code>Batch Selector</code> would freeze the execution pipeline if the user switched browser tabs during an active generation phase.</li><li>ROBUSTNESS: Deployed an active "Heartbeat" loop protocol to guarantee continuous backend-to-frontend synchronization.</li><li>STABILITY: Integrated crash-protection guards to intercept and drop corrupted or incomplete image payloads gracefully inside the canvas preview matrix.</li></ul><h3 id="v1.0.4">v1.0.4</h3><ul><li>NEW NODE: Introduced <code>Aspect Ratio Master</code>. Complete native resolution logic engine supporting architectures like SDXL, Flux, Wan, and LTX.</li><li>ARCHITECTURES: Hardcoded updated 2026 mathematical parameters for Hunyuan 2.1 and Wan 2.2 pipelines.</li><li>STABILITY: Mandated rigorous Mod32/Mod64 latent validation logic to prevent decoding distortions.</li><li>ORGANIZATION: Migrated the <code>Batch Selector</code> module into the unified <code>3R3BOS/Image</code> category menu for pristine workflow mapping.</li></ul><h3 id="v1.0.3">v1.0.3</h3><ul><li>BUGFIX: Fixed a core scaling glitch where the <code>Batch Selector</code> image grid incorrectly processed layout borders. Standardized to strict 'contain' rendering to eliminate clipping on non-square image aspect ratios.</li></ul><h3 id="v1.0.2">v1.0.2</h3><ul><li>NEW NODE: Released <code>Batch Selector</code> as the feature-complete architectural replacement for the legacy <code>Visual Gatekeeper</code> module.</li><li>DESIGN LANGUAGE: Complete UI overhaul across all native widgets—standardizing on a clean, unobtrusive monochrome design philosophy to keep the main graph context-focused.</li></ul>

ComfyUI-3R3BOS-Pack

v1.0.8<ul><li>CRITICAL FIX — Aspect Ratio Master: Latent channel count was hardcoded to <code>4</code> for every architecture. All modern models (Flux, Wan, SD3, CogVideoX, Hunyuan Video, Cosmos, etc.) were generating an incorrect latent shape, causing VAE decode crashes or silent grey outputs. Each architecture now outputs the exact correct channel count: <code>4ch</code> (SDXL/SD1.5), <code>12ch</code> (Mochi 1), <code>16ch</code> (Flux/Wan/SD3/CogVideoX/HiDream/Krea2/AuraFlow/Lumina/Z-Image/OmniGen/Hunyuan Image/Cosmos), <code>32ch</code> (SANA-1.5), <code>128ch</code> (LTX Video).</li><li>FIX — VAE Compression Factor: <code>LTX Video</code> corrected from <code>vae_factor=8</code> to <code>vae_factor=32</code> (DC-VAE). <code>Hunyuan Image 2.1</code> corrected from <code>vae_factor=8</code> to <code>vae_factor=32</code> (32× spatial compression). <code>PixArt Alpha/Sigma</code> corrected to <code>4ch</code> (uses SDXL-compatible VAE, not a 16ch architecture).</li></ul>

Files with SHA256: 2a57d79d521200ba7765adc6e591faf4a4691cd54430958e667dde7ebc00d1ce