Load Image Pro (Path/Image → RGB/RGBA/Mask/Upscale)

Load Image Pro is what you get when someone looks at the stock ComfyUI image loader and says "I want that, plus a mask pipeline, plus an upscaler, in one node." It loads from a path or takes an incoming IMAGE, and gives you RGB, RGBA, a mask with cleanup tools, and an optional upscale pass - all in one box. For batch work where every image needs the same mask treatment and a 2x bump, that's a lot of graph space saved.

How it works

The node has two source paths: a path string, or an existing image input (the image wins if both are given). From there it composes an RGBA version, extracts or builds a mask, runs the mask tools if enabled, and optionally upscales. The mask toolchain is genuinely useful:

• mask_blur - Gaussian blur on the mask (soft edges).
• mask_offset - dilate (positive) or erode (negative) the mask in pixels.
• smooth - morphological open+close to clean up noise.
• fill_holes - flood-fill to close gaps.
• invert_mask - flip it.

The upscale branch uses a real upscale model from ComfyUI's models/upscale_models folder if one is installed and selected, falling back to OpenCV's lanczos/nearest-exact/bilinear/area methods otherwise. The mask is resized to match, so everything stays in sync.

Inputs that matter

• path - where to load from. Empty string with an image input is fine.
• enable_mask_tools - the master switch for the mask cleanup block (the individual tool settings do nothing until this is on).
• enable_upscale + upscale_factor (0.05–8x) + upscale_model + upscale_method - the upscale block. upscale_model lists your installed upscale models; if you have none, it shows (no models found) and the node silently uses the OpenCV method.

Outputs

image (RGB), image_rgba, mask, mask_image, and width/height after any upscaling. Wire the mask into inpainting or compositing; wire the dimensions into anything that needs to know the output size.

How to install

Part of the orion4d/ComfyUI_DAO_master pack:

cd ComfyUI/custom_nodes
git clone https://github.com/orion4d/ComfyUI_DAO_master.git
pip install opencv-python     # cv2 is required for the mask tools and OpenCV upscale

Restart ComfyUI. Model upscaling needs an ESRGAN-style model in ComfyUI/models/upscale_models/ - the pack itself downloads nothing.

Common issues

(no models found) in the upscale model dropdown isn't an error - it just means no upscale model is installed, and the node falls back to OpenCV interpolation. Two settings are silent switches: enable_mask_tools and enable_upscale - the tools do nothing until they're on, which trips up people who set mask_blur and see no change. And remember the image input wins over path: if you wire both and the output looks like the tensor, not the file, that's by design.