ComfyUI-Resolution-Divider

A smart utility node for ComfyUI designed to calculate downscaled resolutions while strictly maintaining the original aspect ratio.

This node is essential for Image-to-Video (I2V) workflows (like Wan 2.2), where using high-resolution input images often leads to OOM (Out of Memory) errors. It helps you find the "sweet spot" (e.g., 480p or 720p) instantly without manual calculations.

🚀 Key Features

Smart Parsing: Accepts various format inputs like 1920 x 1080, 1920x1080, 1920 1080, or 1920, 1080.
Auto-Detect Resolution: Select an image from your input folder or upload one directly on the node. It automatically reads the dimensions.
Real-Time Calculation: The "Live Result" updates instantly via JavaScript as you adjust the slider—no need to queue the prompt to see the numbers.
VRAM Saver: Perfect for downscaling 4K/HD images to manageable sizes for heavy models (Wan 2.2, Kling, etc.) without making the image "squashed" or "stretched".
Clean UI: Optimized widget layout for a smooth user experience.

📦 Installation

Method 1: Manual Clone

Navigate to your ComfyUI custom_nodes directory.
Open a terminal/command prompt.

Run the following command:

git clone https://github.com/CraftBlack/ComfyUI-Resolution-Divider.git

Restart ComfyUI.

Method 2: ComfyUI Manager

Coming soon... (Once registered in the index)

🛠️ Usage

Add the Node:
- Right-click canvas -> Add Node -> Resolution Utils -> Resolution Divider (VRAM Saver).
Input:
- Option A: Type the resolution manually in res_string (e.g., 1024x1024).
- Option B: Click "Upload" or select an image from the dropdown. The node will auto-fill the resolution.
Set Divider:
- Adjust the divider float value.
- 1.0 = Original Size.
- 1.5 = 1.5x smaller (Great for Wan 2.2).
- 2.0 = Half size.
Connect:
- The node outputs a custom RESULT string (for display) and a standard STRING (for info).
- Note: This node is primarily a calculator. You manually use the numbers shown in "Live Result" to set your Latent size, or you can expand this node to output INT width/height in the future.

💡 Why "Divider"?

Unlike "Scale" nodes which usually multiply (upscale), this node divides (downscales).

\text{New Size} = \frac{\text{Original Size}}{\text{Divider}}

Example:

Input: 1280 x 720
Divider: 1.5
Result: 853 x 480 (Perfect 480p safe zone for heavy AI models)

🤝 Contributing

Feel free to submit issues or pull requests if you have ideas for improvements!

📄 License

MIT License. Free to use for everyone.

2.8 KiB Raw Blame History