🖼🔍 Image Details

A small diagnostic node that pulls the facts out of an image and hands them back as usable values instead of something you have to inspect manually - width, height, whether it has transparency, and whether it reads as landscape, portrait, or square. It's not glamorous, but it's the kind of node that lets you build workflows that branch based on what an image actually is, rather than assuming.

Why you'd reach for it

Two common uses. First, plain debugging - you loaded an image from an unfamiliar source and want a quick answer to "wait, what am I actually working with here" without switching to a file browser. Second, and more useful long-term: routing. Feed the ORIENTATION or HAS_TRANSPARENCY output into the pack's own If-Else node and you can branch a workflow automatically - pick a differently-shaped empty latent depending on whether your reference image is portrait or landscape, or route transparent images through an alpha-removal step before they hit a node that can't handle transparency, without you having to check manually every time.

The inputs and outputs that matter

One required input: image_input. Six outputs come back: WIDTH and HEIGHT as integers; HAS_TRANSPARENCY, a boolean - the pack treats RGBA images as having transparency and RGB images as not, so this is a format check rather than a check for whether any pixel is actually see-through; ORIENTATION, one of landscape, portrait, or square as a string; TYPE, the image's format as a string; and ALL, every one of those details bundled into a single string, handy for a quick dump into a Show node when you just want to glance at everything at once rather than wiring up five separate previews.

Installing it

Through ComfyUI Manager: search Bjornulf_custom_nodes, install, restart. Or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/justUmen/Bjornulf_custom_nodes

Install the shared requirements.txt afterward. This is a lightweight inspection node with no heavy dependency of its own - it works fine on a minimal install.

Where people get burned

The main thing to keep straight: HAS_TRANSPARENCY is checking the image's color mode (RGBA vs RGB), not scanning the alpha channel for whether anything is actually transparent. An RGBA image where every pixel's alpha is fully opaque will still report true here - if you loaded it with a node that always outputs RGBA regardless of the source, don't be surprised this comes back positive even when nothing visually looks transparent. If you need to branch on genuine transparency rather than just format, you'd want to inspect the alpha channel's actual values elsewhere in the graph, not rely on this output alone.