IMAGE INFO (JOV) 📚

Every ComfyUI workflow eventually hits the same moment: you're feeding an image into a resize, a crop, or a latent and you need its exact dimensions, but nothing tells you. IMAGE INFO (JOV) 📚 is the node that ends that guessing game. You drop one image in, and out come the batch count, width, height, and channel count as clean numbers you can wire straight into whatever comes next.

It's the simplest node in Jovimetrix, and honestly one of the most useful. Because its outputs are plain INTs, they'll plug into anything that wants a number - a latent upscale-by, a crop size, a controlnet resolution. When you're debugging why an upscale came out the wrong shape, this is the first node you add.

How it works

Jovimetrix stores images as tensors shaped (batch, height, width, channels). IMAGE INFO just reads that shape off the first image in the batch and prints it out the ports. There's no heavy lifting, no model involved - it's instant and it can't break your image. It accepts either an IMAGE or a MASK, which is handy because masks are famously opaque about their dimensions.

The inputs and outputs

One input, six outputs:

• 👾A - the image (or mask) to examine. That's the only thing you set.
• Batch count (INT) - how many images are stacked in the batch. This is the one people forget about until a "single image" node starts choking on 24 frames.
• Width and Height (INTs) - in pixels.
• Channels (INT) - 3 for RGB, 4 for RGBA, 1 for a mask.
• WH (VEC2) and WHC (VEC3) - the same data pre-packaged as vectors, for Jovimetrix's own vector-typed inputs and for ComfyMath-style vector nodes.

A common pattern: read W and H, then feed them into a scale-by-2 node or a latent resize so your upscale is always relative to the actual input. Once you've wired that, the node's job is done - leave it inline, it costs nothing.

Installing it

IMAGE INFO ships with the Jovimetrix pack, so you install the whole pack once:

• ComfyUI Manager - search for "Jovimetrix" and hit install.
• Manual - git clone https://github.com/Amorano/Jovimetrix.git into your custom_nodes folder, then pip install -r requirements.txt.

No model files to download - Jovimetrix is pure computation (numpy, OpenCV, Pillow, plus the author's own cozy_comfyui helper library pulled in from GitHub). It needs ComfyUI 0.1.3+ and a reasonably recent frontend. If you load an old workflow that references the emoji-less "IMAGE INFO (JOV)" name and it doesn't resolve, that's the same node under its newer name.

Gotchas

Not much to trip over here, but two things: if the image port is empty the node returns nothing (it doesn't invent dimensions), and if you're reading values to drive a tiled upscale, remember these are the source dimensions - a 512x512 source with a 2x upscale still reports 512. The node reads what's actually on the wire, which is exactly what you want for debugging. It's the kind of pack-level node where you'll wonder how you lived without it - same energy as Impact Pack's size utilities, just a lot cheaper to run.