Load Images From URL ♾️Mixlab

The README describes this one plainly, and it's worth quoting because it explains the whole point: "conveniently load images from a fixed address on the internet to ensure that default images in the workflow can be executed." This isn't a general-purpose image downloader you'd reach for on its own - it's specifically for making sure a shared or published workflow has something to run on the moment it opens, without waiting on a user to upload a file first.

That framing matters most in the context of Mixlab's Workflow-to-APP feature. When you turn a workflow into a shareable web app, a cold visitor to that app shouldn't hit a blank graph with nothing to sample from - they should see something working immediately, then swap in their own image if they want to. LoadImagesFromURL is how you seed that default: point it at a stable, direct image link, and the app has a working example out of the box.

The url input is a multiline STRING field defaulting to "https://". Multiline plus list-typed outputs is the tell here - this node accepts more than one URL, one per line, and fetches all of them in a batch, which lines up with both outputs being lists: images (IMAGE) and masks (MASK). You get a mask alongside every image whether or not the source has meaningful alpha, since ComfyUI's image pipeline generally expects the pair. There's also an optional seed input, typed as the wildcard * - in a workflow where nodes cache their output between runs, wiring a changing value into seed is a common trick to force a re-fetch instead of ComfyUI reusing a cached result from last time, though the node's schema doesn't document that as its explicit purpose, so treat it as a likely-but-unconfirmed use rather than a guaranteed one.

Install follows the same path as the rest of the pack. Through ComfyUI Manager: search "comfyui-mixlab-nodes," install, restart. By hand:

cd ComfyUI/custom_nodes
git clone https://github.com/shadowcz007/comfyui-mixlab-nodes.git
cd comfyui-mixlab-nodes
install.bat

or pip3 install -r requirements.txt in a venv, or the portable-build pip call. Restart ComfyUI once it's done. No model downloads needed for this one specifically - it's a network fetch, not an inference node.

That network fetch is also where the real gotchas live. First, it needs live internet access from wherever ComfyUI is actually running - fine on a home machine, a genuine constraint if you're running on a locked-down or offline server, since a broken or unreachable URL just fails the run rather than falling back to anything. Second, the URL has to point directly at an image file, not a webpage that happens to display one - a link to an Imgur post page won't work the way a direct .png or .jpg link will. Third, some hosts block hotlinking or require a referer header the node isn't going to send, so an image that opens fine in your browser can still 403 when this node tries to pull it - if a fetch fails, try the file's raw/direct-CDN link instead of the page URL before assuming the node is broken. And per the pack's own general guidance, prefer HTTPS sources; the README calls out HTTPS as a requirement elsewhere in the pack for related upload/network features, and it's the safer default here too.