VRGDG Image Compare

VRGDG Image Compare answers the most common question in any enhancement workflow: is the processed version actually better? It previews two images side by side on the canvas - original vs. enhanced, first pass vs. second pass, model A vs. model B - without saving anything to ComfyUI's output directory. It's built on ComfyUI's PreviewImage, so it slots in wherever you'd put a preview node, and it adds the comparison modes that make judgment calls actually possible.

This pack leans on it hard: the Z-Image upscale workflows, the enhancement passes, the face-repair iterations - all of them produce a "before/after" question, and this node is the author's answer to it. If you're evaluating whether a color-match or grain node helped, or whether a LoRA made things better or just different, this is the node to have on the canvas.

The inputs

• image_a and image_b - the two images. Conventionally A is the original, B is the processed one, but that's up to you.
• mode - the comparison style, and the heart of the node:
  • side_by_side - split layout.
  • slider - a draggable divider; drag to reveal one image under the other. The default, and the best for spotting localized differences.
  • overlay - blend at overlay_opacity, great for alignment and color-shift checks.
  • difference - shows the actual pixel difference, which is brutal and honest.
  • blink - rapidly toggles between A and B at blink_speed (higher = faster blinks). The classic motion-detector trick, and surprisingly effective for spotting flicker.
• batch_index - which frame of a batch to show, if your inputs are videos or batches.
• slider_position / overlay_opacity / blink_speed / show_labels - the mode-specific controls, all with sensible defaults.

The outputs

Both images pass through unchanged (image_a, image_b), so you can drop this node in the middle of a graph to inspect, then continue wiring downstream without breaking the flow. It's marked as an output node, so it's really a viewer.

Installing it

Pack install as always: Manager → search vrgamedev, or:

cd ComfyUI/custom_nodes
git clone https://github.com/vrgamegirl19/comfyui-vrgamedevgirl.git

Restart, hard-refresh. No extra dependencies beyond the pack itself.

Common issues

• Preview looks identical to plain PreviewImage. You're probably in side_by_side with similar images - switch to slider or blink for the actually-useful modes.
• Batch inputs show only one frame. Set batch_index to the frame you care about.
• Difference mode looks like static. That's not a bug - it means the images are very different, or one is misaligned. If you want a meaningful diff, make sure the two inputs are spatially aligned.

There are fancier A/B tools, but few that drop into a graph as cleanly. When someone asks "did that actually help?", this node is the one-word answer.