Load Image Basic (RMBG) 🖼️

ComfyUI's built-in Load Image node only knows how to load files you've uploaded into its input folder. This one also takes a local path or a URL, which sounds minor until you're trying to pull a reference image straight off the web or point at a file sitting somewhere else on disk without copying it into ComfyUI first. That's the whole reason this node exists.

It's a small quality-of-life node, and the RMBG pack ships three flavors of it - Basic (this one), Advanced, and a third variant - because loading images turned out to be a recurring annoyance the author decided to just fix. Basic is the strip-it-down version: give it a source, get back an image and a mask. No resize options, no metadata, no fuss.

How it works

You put a path or URL in the text field and it loads that. Or you leave the text field empty and pick from the image dropdown, which lists files already in ComfyUI's input folder - so it also works as a drop-in replacement for the native loader. If the source image has an alpha channel, you get that as the MASK output for free.

The inputs and outputs that matter

• image_path_or_URL - the star of the show. Paste a local filesystem path or an http(s):// URL. This is what makes the node worth having over the built-in loader.
• image - the dropdown of already-uploaded input files, for when you're not loading from a path. Leave the path field empty to use this.

Outputs are IMAGE, MASK (from the alpha channel, if present), WIDTH, and HEIGHT. The width and height ints are handy to wire straight into a resize or a latent-size node so your graph adapts to whatever you loaded.

How to install it

• ComfyUI Manager: search Comfyui-RMBG, install, restart.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/1038lab/ComfyUI-RMBG, then pip install -r requirements.txt, and restart.

No model, no weights - it loads the moment the pack does.

Common issues

URL loads depend on your machine actually reaching the internet and the URL serving a real image; a link that redirects or requires a login will fail, and the fix is to download the file and point at a local path instead. Local paths are relative to wherever ComfyUI is running, so when in doubt use an absolute path. And if you need more than a plain load - resizing on the way in, choosing which channel becomes the mask, pulling EXIF metadata - that's the Load Image Advanced node in the same pack. Basic is deliberately bare; reach for Advanced when you want the knobs.