Download Image

A one-field node: give it an image URL, get back an IMAGE. Compare it to RiceRoundImageNode, which is RiceRound's version of core LoadImage and picks a file from your local input/ folder - this node does the same conceptual job but fetches over the network instead. That distinction matters more than it sounds like once you're publishing through RiceRound: a workflow running on riceround.online for someone who isn't you can't reach into your local filesystem, so if the published version of your workflow needs to accept an image, a URL is the realistic way it gets one in.

Where it fits in the publish flow

ComfyUI_RiceRound's model is that RiceRound Publish scans your graph for its own Input-category nodes and turns each one into part of the interface it generates. Local file pickers make sense while you're building and testing on your own machine, but they don't generalize to "someone else's browser, hitting your published page." A URL-based input does. So RiceRoundDownloadImageNode is the more production-realistic input node of the two image options - you'd typically use RiceRoundImageNode for quick local iteration and this one, or its mask-inclusive sibling RiceRoundDownloadImageAndMaskNode, for the shape your workflow will actually see once it's live.

The input and output that matter

• image_url (STRING, no default) - the URL to fetch. There's no format guidance in the schema or the README beyond the field being a plain string, so treat it as "point it at a direct image link" and verify by testing.
• Output: value (IMAGE) - the fetched image, wired downstream exactly like any other IMAGE source.

Note there's no mask output here - if your source image needs its alpha channel extracted too, use RiceRoundDownloadImageAndMaskNode instead, which returns both.

Installing it

ComfyUI Manager: search "RiceRound," install, restart. By hand:

cd ComfyUI/custom_nodes
git clone https://github.com/RiceRound/ComfyUI_RiceRound

then restart ComfyUI. The README doesn't list any Python dependencies or model weights for this pack - it's a packaging and network-fetch layer, not a generation pack, so there's nothing heavy to wait on.

Common issues & troubleshooting

The node errors out or returns nothing. Since this makes a live network request, the usual URL-fetch failure modes apply: the URL isn't publicly reachable, it requires authentication the node has no way to provide, or it points to a page rather than a direct image file. None of this is RiceRound-specific - it's the generic cost of a "fetch by URL" node - but it's worth ruling out before assuming the node itself is broken.

Is this safe to leave in a workflow you're about to publish? Yes - it's Input-category, and per the README that's the family of nodes meant to end up in the graph you run through RiceRound Publish. The Output-category nodes are the ones to strip before publishing; this isn't one of them.

You're testing locally and the URL you're pointing at is slow or unreliable. For fast local iteration, temporarily swap in RiceRoundImageNode with a local test file instead, then switch back to Download Image once you're confident the rest of the graph works - no reason to eat network latency on every test run while you're still debugging something unrelated.