ComfyUI-SmartSize

A simple but essential custom node for ComfyUI that automatically detects image dimensions and snaps them to the nearest multiple of 16. This is critical for Flux and SDXL workflows to prevent EinopsError (shape mismatch) and ensure high-quality generation.

🎯 Why SmartSize?

When working with modern AI image generation models like Flux and SDXL, you may encounter errors like:

EinopsError: Error while processing rearrange-reduction pattern "b c (h ph) (w pw) -> b (h w) (c ph pw)"

This happens because these models require image dimensions to be multiples of 16 (or 64 for some configurations). SmartSize automatically handles this for you by:

• 📏 Snapping any input dimensions to the nearest multiple of 16
• 🛡️ Preventing shape mismatch errors in Flux and SDXL workflows
• ⚡ Ensuring optimal performance and quality
• ✅ Supporting a wide range of resolutions (64-8192px)

🚀 Installation

Method 1: ComfyUI Manager (Recommended)

1. Install ComfyUI Manager
2. Search for "SmartSize" in the manager
3. Click Install

Method 2: Manual Installation

1. Navigate to your ComfyUI custom nodes directory:

   cd ComfyUI/custom_nodes/
   

2. Clone this repository:

   git clone https://github.com/xiujiang1987/ComfyUI-SmartSize.git
   

3. Restart ComfyUI

📖 Usage

In ComfyUI Workflow

1. Add the Smart Size (Snap to 16) node to your workflow
2. Connect or input your desired width and height
3. The node outputs the snapped dimensions that are safe to use with Flux/SDXL

Example

Input:

• Width: 1000
• Height: 750

Output:

• Width: 992 (nearest multiple of 16)
• Height: 752 (nearest multiple of 16)

Node Details

Inputs:

• width (INT): Desired width (64-8192, default: 1024)
• height (INT): Desired height (64-8192, default: 1024)

Outputs:

• width (INT): Snapped width (multiple of 16)
• height (INT): Snapped height (multiple of 16)

🔧 How It Works

The node uses a simple but effective algorithm:

snapped_width = round(width / 16) * 16
snapped_height = round(height / 16) * 16

This ensures:

• Values are rounded to the nearest multiple of 16 (not just floor or ceil)
• Minimum dimensions of 64x64 are enforced
• All outputs are compatible with Flux, SDXL, and similar models

💡 Common Use Cases

Flux Workflows

Connect SmartSize before your Flux model nodes to ensure compatible dimensions.

SDXL Workflows

Use SmartSize to prepare dimensions for SDXL's VAE encoding/decoding stages.

ControlNet & LoRA

Prevent dimension mismatches when using ControlNet or LoRA with base models.

Image Resizing

Ensure any custom resolution stays compatible with your model pipeline.

🌟 Popular SDXL/Flux Resolutions

The following resolutions are already multiples of 16 and will pass through unchanged:

• 1024×1024 (1:1 square)
• 1152×896 (9:7 landscape)
• 896×1152 (7:9 portrait)
• 1216×832 (19:13 landscape)
• 832×1216 (13:19 portrait)
• 1344×768 (7:4 wide)
• 768×1344 (4:7 tall)
• 1536×640 (12:5 ultrawide)

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

This project is open source and available under the MIT License.

🙏 Credits

Created to make Flux and SDXL workflows more reliable and user-friendly in ComfyUI.