Resolution to Dimensions (lab)

ComfyUI's latent and image nodes want width and height as separate numbers, but humans think in strings like 1024x1024. Resolution to Dimensions (lab) is the tiny bridge between those two worlds: you type a resolution as text, and it hands you two clean INTs - optionally scaled. It's one of those nodes you don't think about until you're wiring a resolution through a variable, and then it saves you from arithmetic you were definitely going to get wrong.

How it works

The resolution field is parsed with a strict pattern: one or more digits, an x, another run of digits, with optional spaces (1024x1024, 1024 x 1024 both fine). Then each dimension is multiplied by scale_factor and truncated to an integer. If the string doesn't match the pattern at all, the node raises an error and goes red - it doesn't guess, it doesn't warn, it fails.

Inputs: resolution (STRING, default 1024x1024) and scale_factor (FLOAT, default 1). Outputs: width (INT) and height (INT).

Why the scale factor is the useful part

Most nodes of this kind just split the string. The scale_factor is what makes it worth having: need to feed an upscale pass at 2×? Set scale_factor to 2 and the same 1024x1024 string becomes 2048x2048 without a second node. You keep one human-readable resolution in one place, and any node downstream can derive its own dimensions from it. It's a small thing that quietly removes a whole class of "wait, why is my latent 512 wide" bugs.

Wire the outputs anywhere an INT dimension is expected - Empty Latent Image's width/height, an image resize node, a latent upscale - and if the resolution comes from a string you already have elsewhere in the workflow (say, from a config file or a Format String node), this becomes the single conversion point.

Where people get burned

The strict parser. 1024x1024 works; 1024×1024 with a unicode multiplication sign, or 1024,1024, or 1024x1024x4, does not - red node, error message, nothing else. Also remember scale_factor truncates rather than rounds: 0.6 on 1024 gives you 614, not 614.4. And since both outputs are INT, don't expect fractional dimensions no matter what factor you set.

Install

It ships in ComfyLab Pack, so:

cd ComfyUI/custom_nodes
git clone https://github.com/bugltd/ComfyLab-Pack.git
cd ComfyLab-Pack
pip install -r requirements.txt

Or ComfyUI Manager → ComfyLab Pack, then restart. No model files, no special setup - the pack's only notable dependency is opencv-python. For a resolution string that has to survive being passed around and multiplied, this is the tidy way to do it.