Merge Image Batches

Every tile-focused pack ends up with one node that isn't about tiles at all, and this is it: a dead-simple "join these image batches along the batch dimension" utility. If you've split an image, processed the tiles down two different branches, and now want them back in one IMAGE tensor - or you just have a few batches you'd like to concatenate - this does it in a single node. There's nothing clever here, which is kind of the point. ComfyUI's built-in batch tools are fiddly about shapes, and this one exists so you don't have to think.

What it does

You feed it image1 (required) plus up to image2 through image6 (optional). It concatenates whatever is connected along dimension zero and hands you back one IMAGE. Same batch merge ComfyUI's stock nodes do, capped at six inputs, with one opinionated detail: images that don't match image1's height/width/channels are skipped with a warning printed to the console rather than crashing the whole graph.

That's the thing to keep an eye on. A skipped image doesn't raise an error - it just quietly doesn't appear in the output. If your merged batch comes up one image short, check the console for lines starting BatchImages_Tiled: Warning, image... different dimensions. Easy to miss, and the result is a batch where a later node assumes order that isn't there.

When you'd actually use it

Honestly, this node is the least interesting of the five in the pack. Where it earns its keep is as a collector that isn't a collector: after splitting a batch of images into tiles and running a per-tile pass, you might end up with several IMAGE tensors from separate branches that belong in one batch before a final VAE decode or save. Concatenating them here keeps the rest of the graph treating them as a single set. For a one-off merge, ComfyUI's built-in Batch Image / Join Image Batches equivalents do the same job - reach for this one if you're already in a SplitImage workflow and don't want to pull in another pack for one concatenation.

Installing it

Same as the rest of the pack - ComfyUI Manager (search "ComfyUI-SplitImage"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/QuigleyDown/ComfyUI-SplitImage.git

then restart ComfyUI. No dependencies beyond ComfyUI's own torch, no models, no keys.

Gotchas

Two, both mild. First, the six-input ceiling - more than that and you're chaining two of these nodes, which is silly but works. Second, the silent skip on dimension mismatch: make sure the images you merge are the same resolution, or you'll be debugging a missing batch member by squinting at terminal output.