Image Get Size (EmAySee)

Every so often a workflow needs to know how big an image actually is - to compute an upscale target, to check whether a frame matches a model's native resolution, or to decide which resize branch to take. EmAySee Image Get Size does the obvious thing: feed it an image, get back its width and height as integers.

It's a two-output utility, the ComfyUI equivalent of PIL.Image.size. Wire the width and height into any node that takes INT inputs and you can build math around your image's real dimensions - which is the entire point of the thing. It pairs naturally with a math or conditional node: "if height is under 1024, upscale, otherwise pass through."

One caveat on name-dropping: there are several nodes called "image get size" floating around, and ComfyUI core has had its own image-size reporting since 2024. This one's distinguishing feature is mostly that it lives in the EmAySee pack and returns plain INTs with no extra outputs - no batch count, no aspect ratio, no extras. If you only need (width, height) out, this is the smallest hammer that does the job.

How it works

The mechanism is embarrassingly simple, and it's worth knowing the one subtlety. ComfyUI stores images as tensors shaped (batch, height, width, channels) - height comes before width in the shape. The node reads that shape and returns (width, height), so the outputs are named correctly even though the tuple order in the code is (height, width). The source lives in py/EmAySee_MathShit.py alongside the pack's MathExpression node, which is a hint about how these are meant to be used together: get the size, run an expression, feed a resolution.

The inputs that matter

• image - any IMAGE tensor. That's the only input.

Outputs are width and height, both INT, in pixels of the image's spatial dimensions (not the batch size).

Install

Part of the EmAySee pack - a one-author collection of 60+ nodes, no requirements.txt, no model files. Install once:

cd ComfyUI/custom_nodes
git clone https://github.com/EmAySee/ComfyUI_EmAySee_CustomNodes

Restart ComfyUI (or Manager → search "ComfyUI_EmAySee_CustomNodes"). It appears under the plain EmAySee category.

Gotchas

The obvious trap is assuming width is what you'll feed into every downstream node. Latents, images, and text-encoder conditioning all expect dimensions differently - width and height are straightforward pixel counts here, so double-check what the node you're wiring into actually wants (a latent is 8x smaller, for example). And this is a no-support hobby pack; the README closes with "code provided without any guarantees or warranties or support." For a node this trivial the risk is basically nil, but don't build a mission-critical pipeline on a pack whose author explicitly disclaims support without reading the 10 lines of source first.