ComfyUI Extension: image-resize-comfyui

Authored by ussoewwin

Created

Updated

0 stars

Advanced image resizing node for ComfyUI with aspect ratio preservation and mask support

Custom Nodes (0)

    README

    Image Resize for ComfyUI

    This repository is a direct fork and continuation of the archived project palant/image-resize-comfyui.

    • Original creator: Wladimir Palant (MIT License, archived on 2024-05-11 and now read-only).
    • Fork maintainer: ussoewwin/image_resize_comfyui — actively updated for ComfyUI 2024+ releases, with refreshed packaging/docs and regular compatibility testing.

    Like the upstream project, it delivers a powerful ComfyUI node for resizing images without distorting proportions. This fork continues the exact feature set while ensuring it remains drop-in compatible with current ComfyUI builds. Optional mask handling remains fully supported, so masks stay aligned with the resized image. Any future fixes or enhancements will be contributed here while honoring the original design.

    A ComfyUI node titled "Image resize" with inputs pixels and mask_optional, outputs IMAGE and MASK as well as a variety of widgets: action, smaller_side, larger_side, scale_factor, resize_mode, side_ratio, crop_pad_position, pad_feathering

    Features

    • Aspect Ratio Preservation: Resize images while maintaining their original aspect ratio
    • Three Resize Modes: Choose between resize-only, crop-to-ratio, or pad-to-ratio operations
    • Flexible Sizing Options: Use smaller_side, larger_side, or scale_factor for intuitive control
    • Smart Resize Modes: Reduce-only, increase-only, or unrestricted resizing
    • Mask Support: Optional mask input is automatically resized and transformed alongside the image
    • Precision Cropping/Padding: Fine-tune crop and pad positions with pixel-level control
    • Feathering Support: Smooth mask transitions to prevent visible borders during inpainting

    Installation

    Clone this repository into your ComfyUI/custom_nodes directory:

    cd ComfyUI/custom_nodes
git clone https://github.com/ussoewwin/image-resize-comfyui
cd image-resize-comfyui

    Then restart ComfyUI.

    Node Configuration

    action

    Determines how the node processes the image after resizing:

    • resize only: Image is resized while keeping its aspect ratio. The side_ratio setting is ignored.
    • crop to ratio: After resizing, parts of the image are removed to match the target side_ratio. Useful for enforcing specific aspect ratios.
    • pad to ratio: After resizing, transparent padding is added to match the target side_ratio. Original image content is always preserved.

    smaller_side, larger_side, scale_factor

    Only one of these can be active (set to a non-zero value). They determine the target size:

    smaller_side (Integer, 0-8192)

    The smaller dimension of the image will be resized to this value.

    Example: Original image 512×768 px, smaller_side=1024 → Result: 1024×1536 px

    larger_side (Integer, 0-8192)

    The larger dimension of the image will be resized to this value.

    Example: Original image 512×768 px, larger_side=1024 → Result: 683×1024 px

    scale_factor (Float, 0.0-10.0)

    Direct scaling multiplier.

    • < 1.0: Downscaling
    • > 1.0: Upscaling
    • = 0.0: No scaling, only crop/pad as needed

    If all values are zero, the image is not resized but only cropped/padded as necessary.

    resize_mode

    Controls which direction resizing is applied:

    • reduce size only: Images already smaller than the target size will not be resized. Only downscaling occurs.
    • increase size only: Images already larger than the target size will not be resized. Only upscaling occurs.
    • any: Images are always resized to match the target, regardless of direction.

    side_ratio (String, default: "4:3")

    Sets the target aspect ratio when using crop to ratio or pad to ratio actions.

    Format: width:height

    Examples:

    • 4:3 – Standard aspect ratio
    • 16:9 – Widescreen
    • 1:1 – Square
    • 512:768 – Explicit dimensions (when combined with smaller_side=512, always produces 512×768 px)

    crop_pad_position (Float, 0.0-1.0, default: 0.5)

    Controls where cropping or padding occurs:

    • 0.0: All removal/addition on the right/bottom side
    • 0.5: Centered (even on both sides)
    • 1.0: All removal/addition on the left/top side
    • 0.3: 30% on left/top, 70% on right/bottom

    For Cropping: Determines which parts are removed. For Padding: Determines where transparent padding is inserted.

    pad_feathering (Integer, 0-8192, default: 20)

    When padding is applied, this setting gradually expands mask transparency into the original image by the specified number of pixels. This helps avoid visible borders when the padded image is later used in inpainting operations.

    • 0: No feathering (hard edge)
    • 20: Smooth transition over 20 pixels (default)
    • Higher values: Wider feathering range

    Inputs

    • pixels (IMAGE): The input image to resize
    • mask_optional (MASK, optional): An optional mask that will be resized alongside the image

    Outputs

    • IMAGE: The resized (and optionally cropped/padded) image
    • MASK: The resized mask (or generated mask if input was None)

    Usage Examples

    Example 1: Upscale to 1024px minimum dimension

    action: resize only
smaller_side: 1024
resize_mode: any
side_ratio: 4:3 (ignored in resize only mode)

    Example 2: Create square images by cropping

    action: crop to ratio
smaller_side: 512
side_ratio: 1:1
crop_pad_position: 0.5 (center crop)
resize_mode: any

    Example 3: Create 16:9 images by padding

    action: pad to ratio
larger_side: 1024
side_ratio: 16:9
crop_pad_position: 0.5 (center padding)
pad_feathering: 30 (smooth edge transition)
resize_mode: any

    Example 4: Scale to exact size

    action: resize only
smaller_side: 512
scale_factor: 0 (disabled)
resize_mode: any
side_ratio: 512:768 (stored for reference)

    Then set smaller_side=512 and side_ratio=512:768 together for precise dimensions.

    Technical Details

    • Interpolation: Uses bicubic interpolation with anti-aliasing for high-quality resizing
    • GPU Acceleration: Fully compatible with CUDA and CPU fallback
    • Data Format: Works with PyTorch tensors and maintains float32 precision
    • Mask Handling: Masks are resized independently and clamped to [0.0, 1.0] range
    • Validation: Comprehensive input validation with clear error messages

    License

    MIT License - See LICENSE file for details

    Credits

    Created by Wladimir Palant

    Changelog

    v1.0.0

    • Initial release
    • Support for three resize actions (resize, crop, pad)
    • Flexible sizing options (smaller_side, larger_side, scale_factor)
    • Three resize modes (reduce only, increase only, any)
    • Mask support with feathering
    • Comprehensive parameter validation

    Contributing

    Contributions are welcome! Please feel free to submit issues and pull requests.

    Support

    For issues, questions, or suggestions, please open an issue on the GitHub repository.