EA Image Compare
Two images, side by side, captioned — no more tab-flipping
- image_1
- image_2
- comparison
Every LoRA test produces the same problem: you've got a "before" and an "after" and you need to actually look at them next to each other. Tabs in your browser don't cut it at 200% zoom, and stacking two PreviewImage nodes makes you scroll. EA Image Compare stitches both images into a single output with a caption bar under each - so one PreviewImage node, one glance, and the difference is obvious.
This is the pack's core comparison node, and the sibling 3-Way and 4-Way variants are literally this node with more image slots. Learn this one and the others are free.
How it works
It takes two IMAGE inputs (image_1, image_2) and two caption strings, then composites everything into one image using PIL: each input gets a caption bar (default 60px tall, 32px font) drawn on top, and the two captioned panels are pasted side by side with a few pixels of spacing between them. The single composited result comes out of the comparison output - wire that to PreviewImage or any save node.
The parameters are all cosmetic but a few earn their keep:
scale(0.1–2.0) resizes both images proportionally. Crank it below 1.0 to fit a 2K pair on screen; go above 1.0 to pixel-peep. The final composite honors whatever you set, preserving aspect ratio.caption_1/caption_2default toWithout LoRA/With LoRA, which tells you the author's own use case: A/B LoRA testing. Change them to whatever's meaningful to you - weights, prompts, seeds.font_size,caption_height,font_color,background_colorstyle the caption bars. There's also an optionalfontthat pulls from yourcomfyui-kjnodesfonts folder if you have that pack installed; otherwise it falls back to PIL's default bitmap font, which is fine but renders small.
The gotcha that bites everyone
All input images must have identical dimensions - the node raises a ValueError if they don't, and this is the most common error you'll see. Two renders at 1024×576 vs 1024×512 will refuse to compare. Fix it upstream with an image scale/resize node before the compare, or use scale - that only resizes proportionally, so mismatched aspect ratios will still fail. Standardize the size first, then compare.
There's no "export caption text" here - it's a viewing tool. The output is an image, so you can also save it, which is genuinely handy: a single PNG showing your whole A/B test, dropped into a gallery or a Discord post.
Install
Part of the comfyui-ea-nodes pack: ComfyUI Manager → search comfyui-ea-nodes, or git clone https://github.com/ExoticArts/comfyui-ea-nodes into custom_nodes/, then restart and hard-refresh (Ctrl/Cmd+Shift+R). No extra Python dependencies - the pack's requirements.txt is empty and everything it needs ships with ComfyUI.
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| image_1 | IMAGE | — | |
| caption_1 | STRING | Without LoRA | — |
| image_2 | IMAGE | — | |
| caption_2 | STRING | With LoRA | — |
| scale | FLOAT | 1.000.1–2 | — |
| font_size | INT | 3212–200 | — |
| caption_height | INT | 6020–200 | — |
| font_color | STRING | white | — |
| background_color | STRING | black | — |
| spacing | INT | 40–50 | — |
| fontopt | COMBO | default | 1 options: default |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| comparison | IMAGE | — |