Scorpiov Anywhere

You know the drill: you've got one MODEL output that needs to reach a KSampler, a LoRA loader, a second KSampler, and maybe a detailer, and you're dragging wires across the whole canvas like it's 2010. Scorpiov Anywhere exists so you don't have to. Plug a source into it once and any other node in your graph with an empty, type-matching input picks it up automatically - no wire needed on the receiving end.

The name is the whole pitch: anything you connect to it becomes available anywhere. It's a rebuild of the beloved "Anything Everywhere" node from cg-use-everywhere, done independently because that older implementation got flaky on ComfyUI's newer Vue-based frontend. If you've used cg-use-everywhere, the idea will feel instantly familiar; if you haven't, think of it as a wireless broadcast tower for your graph.

How it actually works

Here's the clever part, and it's worth understanding before you build a huge workflow on top of it: this node never runs. It's a JS-only trick. When you hit Queue, the frontend rewrites the serialized graph - adding real links from your source node straight to every matched target, then deleting the Scorpiov Anywhere node entirely. The Python backend never sees it. That's why it's fast, and it's also why broadcast connections never appear as real wires on the canvas. Select the node and a faint dashed visualizer shows you what it's feeding; select a fed node and you see the line back. If you want certainty, check the browser console for skipped-match warnings.

The inputs are dynamic - a single anything_1 slot that accepts literally anything (the ComfyUI * wildcard type), and a fresh empty slot grows every time you fill one. Each connected slot auto-labels itself with its source type (MODEL, CLIP, VAE...).

Matching works on type and name, with priority: an exact name match beats a type-only match. Genuine ties - two CONDITIONING outputs with nothing to disambiguate them - are skipped and logged rather than guessed. That's the safety valve: right-click any input dot → "Rename for Scorpiov matching..." and give it a name like positive so it finds the right target. The engine is also cycle-safe - it won't create a dependency loop through existing connections.

Wiring it in

There are no outputs here, and that's fine - it's a broadcast node. Just wire sources into its inputs and let the targets stay empty. You can also skip the node entirely: right-click any node → "Add Scorpiov Broadcasting" to broadcast its outputs directly.

Gotchas worth knowing

• It's not magic for metadata. Because broadcasts aren't real graph links, the pack's own Image Meta Reader (which walks real links to auto-detect prompts and LoRAs) won't see anything delivered via broadcast. If you need things embedded in your saved PNG, wire them directly.
• The JS rewrite only happens through the normal queue path. Call the API directly and you bypass the rewriter - the Python fallback just no-ops, so nothing crashes, but nothing broadcasts either.
• Right-click renames persist with the saved workflow, so they survive a reload.

Installing

This is one of ten nodes in the scorpiov-nodes pack. Install the pack and you get them all:

cd ComfyUI/custom_nodes
git clone https://github.com/vikramudyawar/scorpiov-nodes

then restart ComfyUI. Or search "scorpiov-nodes" in ComfyUI Manager. No extra Python dependencies - nothing pip-installs, nothing downloads. The JS UI extras are tested on the Vue-based frontend; the older classic frontend should work but is untested, which is exactly the kind of thing to check first if the visualizer or rename menu doesn't appear.

For a single-broadcast convenience, it's maybe overkill - ComfyUI's own shortcuts do fine. Where this earns its keep is the graph with three KSamplers, a couple of LoRA loaders, and a detailer all wanting the same model and clip. One source, one slot, zero spaghetti.