None Image

None Image has no inputs and one output: an IMAGE-typed output that is actually None. It's the "plug in nothing" node - a way to explicitly feed an empty image into nodes that accept one, without leaving a dangling wire or building a fake black image.

Why you'd reach for it

Some ComfyUI nodes have optional image inputs, and some workflows want to conditionally supply an image - pass real pixels in one branch and nothing in the other. That's where a None constant earns its keep: wire it into a Switch or a conditional, and the node downstream gets a None it can check for. TinyBee's own image nodes (like Crop/Grow Image to Bounds or Rect Inside Image) are written to handle None gracefully, so inside this pack it slots right into "is there an image here at all?" logic.

How it works

The entire implementation is return (None,) - the node's function returns Python's None wrapped as a single-value tuple, typed as IMAGE so it fits an IMAGE input slot. There's no tensor, no hidden black frame, no size. If you inspect the value you get None, not an array. That's the point: some code paths genuinely need "no image", and a 1×1 black tensor is not the same thing as None to a node that checks is None.

The inputs that matter

There aren't any. It's a constant node - drop it in, wire the none output where you need it.

Installing it

TinyBee pack: ComfyUI Manager → Install Custom Nodes → search "ComfyUI-TinyBee", install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/TinyBeeman/ComfyUI-TinyBee

Under 🐝TinyBee/Casting. Stdlib only.

Common issues

The honest caveat: not every node downstream handles None gracefully. ComfyUI's type system lets this wire into any IMAGE input, but a node that blindly does tensor.shape on its input will crash on None. So this isn't a universal "skip the image" button - it works where the receiving node is written to check for None (TinyBee's are), and it'll blow up elsewhere. If a node errors with an attribute error on None, that's the downstream code not being None-aware, and you'll want a real dummy image instead. It's a niche tool, but when you're building conditional image logic it's exactly the piece you were missing.