Nodes/ComfyUI_StarNodes/⭐ Star Image Compare
ComfyUI Node

⭐ Star Image Compare

A drag-to-wipe before/after slider, right on the ComfyUI canvas

By Starnodes2024·Created 2 years ago·Updated 2 days ago· 106
⭐ Star Image Compare
  • image_1
  • image_2
  • concat_image
compare_position0.50
caption_image1
caption_image2

Comparing two images in ComfyUI is normally a mug's game - you Preview them side by side and squint across the canvas. Star Image Compare replaces the squinting with an actual draggable wipe slider: one image, one divider, drag back and forth. The author calls it "judge your results like a pro," which is fair, because this is the tool you want when an upscale, a refiner pass, or a filter tweak needs a pixel-level A/B.

How it works

You drop two IMAGE inputs - image_1 and image_2 - into the node, and it renders an interactive comparison view with a vertical divider you can drag. compare_position (0.0–1.0, default 0.5) sets where the slider starts; the tooltip is explicit that 0.0 shows Image 2 and 1.0 shows Image 1, so it's a percentage toward image 1, not a "left/right" toggle. The node also stamps captions (caption_image1, caption_image2) onto the comparison, which is genuinely handy when you're comparing "euler vs ddim" or "before upscale / after."

Here's the part people miss: it's an output node with a real tensor output. The concat_image output is a single composited image built from your two inputs (with the caption bars drawn in), so you can save the comparison itself - that's the thing to grab when you're posting an A/B to show someone a setting difference. It saves the interactive pair to ComfyUI's temp folder so the canvas viewer can serve the wipe view, but the concatenated image is what flows out of the graph.

Inputs and outputs

  • image_1 / image_2 - the pair. Both optional, but a comparison with one missing is just a preview.
  • compare_position - initial slider position; you can also leave it and just drag in the viewer.
  • caption_image1 / caption_image2 - labels drawn onto the result. Use them.
  • concat_image (output) - the labeled side-by-side, ready for Save Image.

Where it fits

After an upscale chain (SeedVR2 vs a plain ESRGAN, say) it's the fastest honest check, because you can wipe right at the edge of a face or a fabric weave where upscalers diverge. Same for refiner and filter tweaks - the KB's advice to change one variable at a time on a fixed seed is much easier to follow when the comparison is one drag away.

Installing

Part of the StarNodes pack - install Starnodes via ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Starnodes2024/ComfyUI_StarNodes
cd ComfyUI_StarNodes
pip install -r requirements.txt

Restart, then search star on the canvas. No models needed.

Common issues

If the two images have wildly different dimensions, the node resizes to make them comparable - don't be surprised the concat output isn't your original resolution. And the wipe view only shows up when you run the workflow (it renders to temp); a fresh node before the first run looks empty until then. If the slider feels like it's not there, check your ComfyUI frontend is up to date - older frontends don't render the interactive viewer.

Category⭐StarNodes/Image And Latent

Inputs (5)

NameTypeDefaultDescription
image_1optIMAGE
image_2optIMAGE
compare_positionoptFLOAT0.500–1Initial slider position. 0.0 = Image 2, 1.0 = Image 1.
caption_image1optSTRING
caption_image2optSTRING

Outputs (1)

NameTypeDescription
concat_imageIMAGE