Load Img Batch From URLs (EZ)

Load Img Batch From URLs (EZ) is the batch version of the pack's single-URL loader: paste a list of image URLs, one per line, and get a single batched IMAGE tensor out the other side - ready to pipe into a VAE encode, a batch of img2img passes, or a training dataset build.

Why you'd reach for it

ComfyUI's built-in Load Image only reads from disk, and its Load Image From Web only does one URL at a time. When you're pulling a set of references from a remote service - a gallery, an API, a dataset export - you don't want to download them manually into input/ just so ComfyUI will notice them. This node skips that whole step. Paste, run, done.

How it works

The urls field is a multiline string, and the node splits it on newlines, loads every URL through the same loader as the single-URL node (which handles regular HTTP, data: URIs, and s3:// URIs), and stacks them into one tensor. One quirk worth knowing: if the images aren't all the same resolution, the node upscales the mismatched ones with bilinear resampling to match the first image's dimensions before stacking. That keeps the batch shape valid, but it can soften images that needed resizing.

The inputs and outputs

urls (STRING, multiline) - one URL per line. This is the only input.
images (IMAGE, output) - the whole batch in a single tensor, first image's dimensions win.

Installing it

From ComfyUI Manager, search easy-comfy-nodes, or:

cd ComfyUI/custom_nodes
git clone https://github.com/wmatson/easy-comfy-nodes
# restart ComfyUI

Where people get burned

Blank lines are a crash, not a skip. The code splits on newlines and loads every entry - an empty line becomes a URL that fails to load. Make sure the field ends with a real URL, no trailing newline.
Mixed resolutions get stretched. If your URLs point at images of different sizes, the smaller ones are upscaled to match the largest (the first one). You'll get a valid batch with soft corners. If fidelity matters, pre-normalize your source images.
Same 5-second timeout as the single loader. A slow server means requests.get(timeout=5) bails, and the whole node fails. If you're hitting a laggy CDN, retry or pre-download.
No mask output. The single-URL node gives you a mask; the batch node gives you just images. If you need alpha per image, load them individually.

It's a small node, but it's the kind of small node that saves you twenty minutes of manual downloading every time you work with remote image sets.