ImageSizeInfo

ImageSizeInfo is about as small as a node gets: feed it an image, get its width and height back as numbers, and the image itself passes straight through unchanged. It exists for one reason - ComfyUI doesn't otherwise give you an easy way to read an image's dimensions out as INTs you can wire elsewhere, and there are plenty of moments where you want that. Matching an EmptyLatentImage to a reference photo's aspect ratio, sizing an ImageAddText caption box to fit the source image, or just confirming what size something actually came out at after a chain of resizes you're no longer sure about.

How it works

There's no processing here beyond reading the tensor's shape. You splice it into your graph between two nodes that already pass an image along - it doesn't alter pixels, add a mask, or touch anything about the image itself. It's a tap, not a filter.

The inputs and outputs that matter

• image - the only input, required.

Three outputs: IMAGE is the same image you fed in, unchanged, so you can keep the pipe going without a detour. width and height are plain INTs you can wire into anything that wants a size - Chibi's own Loader or ImageAddText width/height fields, an EmptyLatentImage, a math node, or a comparison against a fixed value if you're branching logic on image size.

The node is flagged is_output_node, which means ComfyUI will always execute it even if width and height aren't wired anywhere downstream - so you can drop it into a graph purely to glance at the numbers in its UI overlay without needing to consume the outputs at all.

How to install it

• ComfyUI Manager: search "ComfyUI-Chibi-Nodes", install, restart.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/chibiace/ComfyUI-Chibi-Nodes/, restart.

No models, no dependencies - it's a couple of lines of logic reading a tensor's shape. It shows up under Chibi-Nodes/Image once installed.

For the pack as a whole, keep the author's own framing in mind: the README describes these as "experimental nodes" and jokes that version 2 shipped "with more bugs." A node this simple has essentially nothing to break, but it's a fair reminder that this is a small personal utility collection rather than something with dedicated maintenance behind it.

Common issues

Honestly, there isn't much to go wrong. The one thing worth knowing: because the IMAGE output is a passthrough rather than a copy in any meaningful sense, splicing this node in mid-graph shouldn't change your results at all - if it does, the bug is somewhere else in your workflow, not here. And if you're expecting width/height to update live as you tweak an upstream resize node, remember ComfyUI only re-runs a node (and this one) when its inputs actually change and you queue a new prompt - it won't recompute on its own just because you're staring at the graph.