心宝❤智能拼图

心宝❤智能拼图 (XinbaoSmartGrid) does exactly one thing: it takes up to four input images, resizes them so the cells line up, and merges them into a single grid image. Output is one IMAGE. That's the whole job, and it's a job you'll actually want.

It ships in the comfyui-banana-li-linux pack from Li Xinbao (李心宝), the e-commerce-focused author best known for the pack's 心宝❤Banana Gemini API node. The grid node is a quiet utility in a pack that's mostly about paid cloud generation - which makes it notable for what it isn't: no API key, no network call, no cost. It's a local, deterministic image op that runs on CPU and finishes instantly.

Why you'd reach for it

The classic use is a comparison contact sheet. Generate four variants of a product shot with a generator's batch output, drop them into image_1 through image_4, and get one image that shows them side by side. You're no longer tabbing between previews trying to remember which one had the better fabric texture - it's all in front of you, ready to save or drop into a document.

The pack's author works in e-commerce, where "generate 4, pick 1, deliver" is a daily rhythm. The node normalizes each input's dimensions as it builds the grid, so you don't get ragged edges where one cell is 1024 wide and the next is 512. The trade-off: it resizes, so don't use this as your final rendering step. Decide here, re-generate or upscale upstream.

The inputs that matter

All four image inputs are optional - image_1, image_2, image_3, image_4. Plug in whatever you have; a two-image grid works fine. The single IMAGE output feeds a preview, save, or anything downstream that takes an image.

Installing

Same as the rest of the pack:

cd ComfyUI/custom_nodes
git clone https://github.com/98624017/comfyui-banana-li-linux

Restart ComfyUI, or install via ComfyUI Manager by searching "banana-li-linux". Note this is the Linux build - Windows uses comfyui-banana-li, macOS comfyui-banana-li-mac. The pack pulls in opencv-contrib-python and friends on install.

Gotchas

There aren't many, because there's barely anything to configure. The node shows as 心宝❤智能拼图 in the node menu - search "心宝" or "拼图". One thing to keep in mind: because it resizes to fit the grid, don't expect native pixel fidelity in the merged output. And it's a comparison tool, not an assembly tool - if what you actually need is a long vertical detail-page image, the pack has a sibling node (心宝❤详情拼合) for that, and this one isn't it.

Simple, free, and does what it says. Sometimes that's exactly the node you need.