Side By Side Compare (RMBG) 🖼️🖼️

Side By Side Compare stitches two or three images into one labeled comparison strip. Feed it a before and an after, give each a caption, and it hands back a single image with them laid out next to each other and titled. It's a small utility, but a genuinely useful one when your whole reason for existing is comparing background removers - which, in this pack, it often is. "RMBG-2.0 vs INSPYRENET vs BiRefNet on the same photo" becomes one shareable picture instead of three separate outputs you have to eyeball across tabs.

There's no model here and nothing to download - it's pure image composition. It takes your images, normalizes their sizes so they line up, draws each one with its label, and glues them together. That's it. The value is entirely in convenience: it saves you doing this by hand in an image editor every time you want to show a result.

The inputs: image1, image2, and image3 are the pictures (all optional - feed two if you only want a pair, and the third slot is simply skipped). text1, text2, and text3 are the captions drawn over each, defaulting to "Image 1", etc. - set these to whatever you're comparing ("Original", "BiRefNet", "SDMatte"). size_base decides how the differently-sized inputs get reconciled: largest scales everything up to the biggest, smallest down to the smallest, or you can pin the layout to image1/image2/image3's dimensions. text_color (default black) and bg_color (default white) style the labels and the surrounding canvas. The single output is IMAGE - the finished comparison strip, ready to preview, save, or post.

Install rides along with the pack: "Comfyui-RMBG" in ComfyUI Manager, or cd ComfyUI/custom_nodes && git clone https://github.com/1038lab/ComfyUI-RMBG, then pip install -r requirements.txt and restart. Nothing extra to fetch - it's just code.

Practical notes. size_base is the setting that trips people up: if your two inputs are wildly different resolutions and the result looks lopsided or blurry, switch the base - largest keeps the most detail (it upscales the smaller one), smallest keeps files lean (it downscales the larger). For transparent cutouts, remember that a PNG with alpha will composite against bg_color, so pick a background that actually reveals fringing if you're comparing edge quality - a mid-grey or a saturated color shows haloing that white hides. That's a real trick when you're judging background removers: a bad cutout looks fine on white and ugly on red.

Where it fits: this is a presentation node, not part of your generation pipeline. It's the thing you tack on the end when you want to document or share a comparison - model A versus model B, before versus after an edit, three settings side by side. It's not going to win any search traffic (the click-through here is basically zero, which fits - nobody googles "how do I compare two images"), but if you're the kind of person building these background-removal graphs and posting the results, it quietly saves you a trip to Photoshop every single time.