Multi Image Batch (Soze)

Stock ComfyUI only gives you a two-input "Batch Images" node, so combining outputs from three, four, or six different branches means chaining several of them together - a small annoyance that adds up once your graph has multiple generation paths feeding into one comparison or save step. Multi Image Batch just gives you six slots up front and combines whichever ones you've actually connected.

How it works

All six inputs are optional, and that's the entire design: connect images1 through images6 in whatever combination you need - two, four, all six - and it concatenates them into a single IMAGE batch on the image output. Leave the rest disconnected and they're skipped, no dummy inputs required.

As with any batch-combining operation in ComfyUI, the images going in need to share the same resolution. This isn't specific to this node - it's how batching works everywhere in the graph - but it's the one thing that trips people up when merging outputs from different sources (say, one branch upscaled and another not) that happen to not match dimensions.

The inputs and outputs that matter

• images1–images6 (all optional, IMAGE) - connect only what you need.
• image (output, IMAGE) - the combined batch, in the order you wired the inputs (1 through 6).

That ordering matters if anything downstream cares about sequence - feeding a labeled comparison grid, for instance, where you want a specific batch to land in a specific position.

How to install it

ComfyUI Manager → search "ComfyUI_Soze" → install → restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/SozeInc/ComfyUI_Soze.git
pip install -r ComfyUI_Soze/requirements.txt

then restart. No model files, no external dependency - this is a pure tensor-concat utility.

Common issues & troubleshooting

Mismatched resolutions are the number-one failure mode for any batch-combine node, and this one is no exception. If a branch is even a few pixels off from the others (a resize node upstream with a slightly different rounding rule, say), the combine will error rather than silently pad or crop. Run each branch through a consistent resize step before this node if you're not certain they match.

If you're building a comparison workflow - collecting variations from several separate branches and then laying them out as a grid - this is the node that assembles the batch first; XY Image, elsewhere in this pack, is the one that turns an assembled batch into a labeled grid. They're built to be used together: combine here, visualize there.

Beyond that, there's not much to break. If an output looks empty or short, double-check which slots you actually wired - it's easy to think you connected all six and actually left one dangling.