🐯 YFG Image Side by Side

The one image-comparison task you do a hundred times a day in ComfyUI - "did the seed change actually help?" - is annoyingly manual. You can eyeball two previews, or wire up an Image Comparer if you've got one of the big utility packs installed. This node is the lightweight version: take two images, paste them into one canvas, slap a text label on each half, and hand the result to a SaveImage node. That's the whole job, and it does it well.

It's part of the YFG Comical pack (author Manny Gonzalez, a one-person utility-node collection that mostly stays quiet on Reddit - don't expect much chatter about it, the README and source are the docs). It ships with a couple of dozen other tools, so you usually end up with it as a side effect of installing something else you actually wanted.

What the two modes do

The mode input is the one that matters. Split scales both images into a single canvas with a dividing line between them - classic before/after, both halves the same size. Side-by-Side keeps each image at its own dimensions and lays them next to each other with a gap. Same idea, different feel: Split is what you want for A/B, Side-by-Side is better when the two images have very different resolutions and you don't want either squashed.

display just controls what the node shows in its preview: the combined image, or one source image or the other. Handy when you want the node's thumbnail to show something specific.

The inputs that actually matter

• image1 and image2 - the two IMAGE tensors you're comparing. Only these are strictly required.
• image_label - when on, draws a text tag on each half using image1_label / image2_label (defaults "image1" / "image2"). Rename those to "seed 1337" vs "seed 1338" and the output is self-documenting.
• font_size, font_color, font_file - cosmetic. Default font_file is micross.ttf, a Windows font. On Linux or Mac that file doesn't exist, and the node quietly falls back to PIL's built-in default font - which is tiny and ugly. If your labels look like ants, set font_file to any TTF on your system and they'll render properly.

Output

Single Combined Image output (IMAGE). Since this is an output node it also previews directly in the graph - wire it into SaveImage (or just let the preview serve as your contact sheet) and you're done.

Installing

The pack install covers this node. In ComfyUI Manager, search "YFG Comical" and install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/gonzalu/ComfyUI_YFG_Comical

Then restart ComfyUI. Dependencies are light. One quirk: the pack imports cv2 and matplotlib at module load in a couple of its files, and those aren't in pyproject.toml. If the whole pack fails to import after install, pip install opencv-python matplotlib and restart - most ComfyUI setups already have both.

Where people get burned

The font fallback above is the classic one. The other is a subtle trap: because both images get resized to fit, Split mode destroys the resolution difference between your two images. If you're comparing a 512px render against its 2x upscale, Split will scale the big one down so they match - which defeats the point of comparing detail. Use Side-by-Side for that comparison, or compare in the graph previews instead. For quick sanity checks inside a workflow - "did adding this LoRA change anything?" - this is the node you reach for.