Karthik Acharya
f129ffb1e2
Expanded the WanVideoAutoImgResize node to support industry-standard resolution presets, multiple resize modes (Keep Proportion, Pad, Crop, Stretch), smart orientation detection and priority, flexible padding options, and multiple interpolation methods. Updated the README with detailed usage instructions, features, and workflow examples. Improved performance with GPU acceleration and efficient memory handling for video and batch processing.
2.4 KiB
2.4 KiB
ComfyUI WanVideo Auto Resize
A high-performance ComfyUI custom node for intelligent image resizing with industry-standard presets and smart orientation detection. Optimized for video workflows and batch processing.
Features
- Industry Standard Presets: 480p, 576p, 720p, 1080p resolutions
- Smart Orientation Detection: Automatic portrait/landscape/square detection with improved ranges
- Multiple Resize Modes: Keep Proportion, Pad, Crop, Stretch
- Quality Options: Multiple interpolation methods for different use cases
- GPU Accelerated: CUDA optimized with automatic mixed precision
- Flexible Padding: Black, White, Gray, or Blur edge options
Installation
-
Clone into your ComfyUI custom_nodes directory:
cd ComfyUI/custom_nodes git clone https://github.com/your-username/ComfyUI-WanVideoAutoResize -
Restart ComfyUI
Usage
Input Parameters
- Preset Resolution: Choose from standard resolutions or Custom
- Resize Mode: How to handle aspect ratio differences
- Orientation Priority: Auto-detect or force specific orientation
- Interpolation: Choose quality vs speed
Bilinear: Good balance of speed and quality (default)Bicubic: Higher quality, slower processingNearest: Maximum speed, good for pixel artLanczos: High-quality downscaling
Outputs
- image: Processed image tensor
- width/height: Final dimensions
- orientation: Detected orientation
- aspect_ratio: Final aspect ratio
- resize_info: Processing details
Orientation Detection
Improved range detection handles near-square images better:
- Square: Aspect ratio 0.85-1.18 (e.g., 410×390 = 1.05)
- Landscape: Aspect ratio > 1.18
- Portrait: Aspect ratio < 0.85
Performance
- GPU Optimization: Automatic CUDA acceleration when available
- Memory Efficient: Smart memory management for large images
- Fast Interpolation: Nearest neighbor for maximum speed
- Quality Interpolation: Bicubic and Lanczos for best results
Example Workflows
Fast Video Processing
Preset: 720p
Mode: Keep Proportion
Interpolation: Nearest
Orientation: Auto Detect
High-Quality Output
Preset: 1080p
Mode: Pad
Interpolation: Bicubic
Orientation: Auto Detect
License
MIT License - see LICENSE file for details.