ComfyUI-HM-Utilities

A collection of custom nodes for ComfyUI that provides useful image processing tools.

Nodes

AutoCropImage

The AutoCropImage node automatically crops images to remove transparent areas by finding the minimum bounding box around non-transparent pixels.

Functionality

• Input: IMAGE (supports batch processing)
• Output: IMAGE (batch of cropped images)

Specifications

The node performs the following operations on input images:

1. Alpha Channel Analysis: For each image in the batch, analyzes the alpha channel (transparency) values

   • For RGBA images: Uses the alpha channel (4th channel) to detect transparent pixels
   • For RGB images: Treats all pixels as fully opaque (alpha = 1.0)

2. Bounding Box Calculation: Finds the minimum bounding box that contains all non-transparent pixels (alpha > 0)

   • Calculates the minimum and maximum X and Y coordinates of non-transparent pixels
   • Each image in the batch gets its own individual bounding box

3. Image Cropping: Crops each image to its calculated bounding box dimensions

   • Creates a new image with the same dimensions as the bounding box
   • Preserves the original image data within the cropped area

4. Output: Returns a batch tensor of cropped images, where all images are padded to the same dimensions (the maximum dimensions among all cropped images)

Usage

1. Connect an image source to the image input pin
2. The node will automatically process all images in the batch
3. Connect the image output pin to any node that accepts IMAGE input (e.g., PreviewImage, SaveImage)

Features

• Batch Processing: Handles multiple images simultaneously with individual processing for each image
• Alpha Channel Support: Works with both RGBA and RGB images
• Automatic Sizing: Each output image is automatically sized to its optimal dimensions
• Transparency Handling: Properly handles transparent and semi-transparent pixels

Category

image/processing

NormalizeImageWithRectangle ⚠️ Development Status

This node is currently under development and testing. Features and behavior may change in future versions.

The NormalizeImageWithRectangle node resizes images to fit within a specified rectangular area while preserving aspect ratio and adding transparent padding when necessary.

Functionality

• Input: IMAGE (supports batch processing)
• Parameters:
  • width: Target width (1-8192, default: 512)
  • height: Target height (1-8192, default: 512)
• Output: IMAGE (normalized images with exact target dimensions)

Specifications

The node performs the following operations on input images:

1. Aspect Ratio Preservation: Calculates the optimal scaling factor to fit the image within the target rectangle while maintaining the original aspect ratio

2. Resize Operation: Resizes the image using bilinear interpolation to the calculated dimensions

3. Padding with Transparency: Adds transparent padding around the resized image to achieve the exact target dimensions

   • For RGB input images: Converts to RGBA and adds transparent padding (alpha = 0)
   • For RGBA input images: Preserves existing alpha channel and uses transparent padding

4. Centering: Centers the resized image within the target rectangle

Usage

1. Connect an image source to the image input pin
2. Set the desired width and height parameters
3. Connect the image output pin to any node that accepts IMAGE input

Features

• Letterboxing/Pillarboxing: Automatically adds transparent bars when aspect ratios don't match
• Batch Processing: Handles multiple images simultaneously
• Alpha Channel Handling: Properly manages transparency for both RGB and RGBA inputs
• Precise Dimensions: Output images always match the specified width and height exactly

Category

image/processing

(c) hiderminer, 2025.