GRAnySelector

This is a graph-organization utility rather than an image or text tool - it takes up to ten optional inputs of any type (ANY_1 through ANY_10, all typed as wildcard *) and forwards one of them as a single output, also typed *. The use case is the same reason nodes like this exist in most utility node packs: you've got two or three possible upstream sources that might feed a given downstream input - a LoRA-baked model versus a fresh checkpoint, a generated image versus an uploaded one - and instead of rewiring the downstream node's input every time you switch which source you're using, you keep both wired into an ANY slot on this node and let it forward the live one.

How it works

Connect whatever you want to pass through into one of the ten ANY_ slots, and it comes out the output port unchanged - no type conversion, no transformation, it's a pure passthrough. Because the type is a wildcard, it'll accept IMAGE, MODEL, CONDITIONING, LATENT, or anything else ComfyUI can wire between two nodes.

One thing worth being upfront about: the schema doesn't expose any explicit index or priority control to say which slot wins if you connect more than one at once. The safest way to use this node is exactly one ANY_ slot connected at a time, swapping which one is wired rather than leaving several connected simultaneously and hoping for predictable behavior. If you need genuine "pick input 3 of 10" logic with an explicit selector, this isn't that node - it reads as a passthrough/router, not a switch with a control input.

The inputs and outputs that matter

• ANY_1 through ANY_10 - all optional, all wildcard type. Wire in whichever single source you want active.
• output - wildcard type, forwards whatever's connected.

How to install it

Via ComfyUI Manager, search GraftingRayman. Manual install:

cd ComfyUI/custom_nodes
git clone https://github.com/GraftingRayman/ComfyUI_GraftingRayman

Restart afterward. This pack needs OpenAI's CLIP installed as a separate step or nothing imports, even for a utility node like this one that has nothing to do with vision models - the whole pack shares one Python module. Portable ComfyUI: .\python_embeded\python.exe -m pip install git+https://github.com/openai/CLIP.git. System Python: pip install git+https://github.com/openai/CLIP.git.

Common issues & troubleshooting

Pack fails to import. Check your startup console for the CLIP error and run the pip install above - it's the same root cause as every other node in this pack.

Output isn't what you expect when multiple slots are wired. As noted above, this node doesn't document a clear priority order for multiple simultaneous connections. Wire exactly one ANY_ slot at a time and swap connections when you want to change source, rather than relying on a specific slot "winning."

Downstream node errors on type mismatch. Because this is a pure wildcard passthrough, it won't catch a type mismatch for you - if you plug an IMAGE into it and wire the output into something expecting CONDITIONING, that's still going to fail downstream exactly like it would without this node in between. It routes, it doesn't validate.

You actually need index-based switching, not passthrough. If your workflow needs "run input 1 this time, input 2 next time" with an explicit control, look at a dedicated switch node elsewhere in your node library - this one's job is routing a live connection through, not selecting by index.