Image Collage

The workflow you just spent an hour tuning generates one image at a time. To compare a before and after, or show three LoRA variants side by side, you'd normally save everything and paste them into an editor. Image Collage is the node that skips the editor: it takes up to eight images and merges them into a single canvas, horizontally or vertically, with your choice of spacing between them. Perfect for A/B comparisons, mood boards, or turning a palette swatch strip into one image you can eyeball at a glance.

How it works

Each of the eight inputs is optional - connect one, three, or all eight; the node walks them in order and only picks up the ones that are actually wired. It takes the first frame of each image batch (so a multi-frame batch collapses to its first image), converts to PIL, and arranges them along your chosen axis. Horizontal layout sums the widths and uses the tallest image's height; vertical layout does the reverse. Images that are smaller than their neighbors get centered against the bigger ones, and the background is white.

Two behavior details to know before they surprise you:

• Spacing applies between images, not at the edges. Set spacing to 20 and you get a 20px gutter between images but nothing added on the outside ends.
• No inputs connected = hard error. This is the one case where it won't degrade gracefully - it raises "requires at least one image," which is honestly the right behavior. Better a clear error than a mystery.

The inputs that matter

• direction - horizontal or vertical, default horizontal.
• spacing - integer, 0–512, default 0. The gap in pixels between images.
• image_1 through image_8 - all optional IMAGE inputs. This is the short list; that's it.

Output is a single merged IMAGE, which you can wire straight into a Save Image node or feed into further processing.

Installing it

Same as the rest of the pack: ComfyUI Manager, search "Colormaster," install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/orion4d/ComfyUI_colormaster
pip install -r requirements.txt   # Pillow, torch, numpy - you already have them

Restart, done. Nothing heavy to download.

Common issues

The batch-collapse behavior is the main gotcha: if you feed it a batch of four generated images, you get the first frame, not a four-up collage. That catches people who expect it to behave like batch-stack nodes. Relatedly, if your images have wildly different dimensions, the white background plus centering is fine, but mismatched aspect ratios can leave awkward empty space - set your batch sizes up front if you want a clean strip. And if you genuinely want to tile a whole batch, you're better off with a dedicated "make image batch" node; this one is for when you have a handful of specific images and want them in one row. For that, it's the least-fuss option in the pack.