Fat Mex Image Compare
Before and after, without squinting
- image_a
- image_b
- comparison
When you're A/B-ing a prompt, a denoise value, or a before/after face swap, the fastest way to see the difference is two images side by side. This node takes image_a and image_b, stitches them into one image with a small dark gap between them, and hands you a single comparison you can preview and save. It's a two-minute utility, and it's the kind of node that's trivial to build and still worth having around.
The output is a comparison IMAGE. Both inputs take the first image of their batch, so you can pipe in batch outputs without ceremony.
How it works
The mechanism is honest and simple:
- It matches the two images' heights - the shorter one gets resized up to match the taller (Lanczos), so the seam is clean.
- It inserts a
gap(default 4px) of dark gray between them. - It concatenates left-to-right:
image_aon the left,image_bon the right.
That's the whole node. label_a / label_b exist as inputs - one honest heads-up: in the current build they're accepted but not drawn. You get the stitch, not the text overlay. Don't expect "Before"/"After" rendered on the output.
When to reach for it
It's the glue for the pack's own workflow style - compare the quick vs quality face swap, compare two denoise settings on the Reference Sampler, compare presets before committing to one. It also makes a nice checkpoint in a chain: save the comparison, keep generating, and you've got a visual history without hunting through your output folder.
It's not a zoom-capable diff tool - for pixel-peeping two crops, you'll still want something with a magnifier. This is for "which of these two is better" at a glance, and it's perfect at that.
Install
cd ComfyUI/custom_nodes
git clone https://github.com/FatMex/ComfyUI-FatMex-Nodes.git FatMex-Nodes
pip install -r FatMex-Nodes/requirements.txt
Or ComfyUI Manager → search "Fat Mex Nodes" → Install → restart. No models, no downloads.
Where people get burned
- Labels don't show up. You set
label_ato "Before" and it's not on the image. As noted, that's the current build - the stitch is all you get. Works fine; just know the fields are decorative right now. - "It's blurry." The height-matching resize upscales the smaller image, and upscaling a small image to match a tall one softens it. That's inherent to stitching images of different sizes, not a bug - wire in the Fat Mex Upscaler first if you need both at the same resolution.
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| image_a | IMAGE | First image (left side). | |
| image_b | IMAGE | Second image (right side). | |
| label_aopt | STRING | Before | Label for image A. |
| label_bopt | STRING | After | Label for image B. |
| gapopt | INT | 40–50 | Gap between images in pixels. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| comparison | IMAGE | — |