Dickson Load Image

ComfyUI's built-in Load Image is fine, but it's stingy with information: it hands you an IMAGE and a MASK, and that's it. Dickson Load Image is a fork of that node that also tells you the file name and the pixel dimensions - and gives them to you as plain values you can plug into other nodes. If you've ever wished you could feed a file name into a Save Image node's filename field, or grab the width and height as integers for a latent resize, this is the node you actually want.

What you get that the stock node doesn't

Core Load Image gives you two outputs. This one gives you six:

• IMAGE and MASK - identical behavior to the built-in node, including the mask quirk: if the file has no alpha channel you get a 64×64 mask of zeros, which is standard Comfy behavior, not a bug.
• FILENAME - the file's base name without the extension, as a string.
• IMAGE WIDTH INT and IMAGE HEIGHT INT - the pixel dimensions as integers.
• IMAGE INFO - a single string like Width: 1024\nHeight: 1024 (added in v1.0.6).

The whole point is the extra outputs. IMAGE WIDTH INT and IMAGE HEIGHT INT are what you wire into anything that takes an integer size - an upscaler, a latent scale, a conditioning resize - so you're not hardcoding dimensions that change every time you swap the input image. FILENAME plugs into a Save Image filename pattern or into a text node for a prompt that references the source. IMAGE INFO is mostly for display or debugging.

How it works

Under the hood it's the core ComfyUI LoadImage code with extra plumbing bolted on. It reads from your input folder (the dropdown lists files there, with an upload button), applies EXIF orientation, converts to RGB, and returns a batch tensor. The filename is derived by stripping the extension, and the dimensions come off the first frame - so a multi-frame GIF reports the first frame's size. Nothing fancy, which is the point: it's a loader, it should be boring.

Installing it

Same pack as everything here: dicksondickson/ComfyUI-Dickson-Nodes. ComfyUI Manager → search dicksondickson, or:

cd ComfyUI/custom_nodes
git clone https://github.com/dicksondickson/ComfyUI-Dickson-Nodes

Restart after. The only Python requirements are Pillow and opencv-python-headless, which Manager installs automatically.

Gotchas

It's a fork of a core node, so if ComfyUI changes how its Load Image behaves in a future release, this one can lag behind - it's not a core update, it's a snapshot. And if the exact filename matters to you, note the output strips the extension: photo_v2.png comes out as photo_v2. That's usually what you want for naming, but it's worth knowing before you build a filename scheme around it. All in all, it's the rare convenience fork that earns its place - a stock node, plus exactly the outputs you kept wishing it had.