Icy ImageBatch

IcyImageBatch is the ComfyUI-IcyHider wrapper of core ImageBatch - same node, renamed class so the privacy extension can hide it. Nothing new here functionally; if you don't use the hiding, grab the core node.

What it does: takes image1 and image2 (both IMAGE) and concatenates them into a single IMAGE with a batch of two. That's the whole job - no blending, no layout, no cropping. The two inputs have to be the same dimensions and the same number of channels, because all this node does is stack them along the batch axis so downstream nodes see "two images in one tensor" instead of "two separate images."

Why you'd reach for it

ComfyUI is happiest when images travel in batches. Some things you can only do to a batch:

• Batch sampling. Concatenate two reference images and feed them to a single pipeline that processes the whole batch in one queue run - each image gets its own result, and you get both back out of the same VAEDecode.
• Batch conditioning. IP-Adapter-style and some ControlNet/refinement paths operate per-batch; feeding them a combined tensor keeps the graph small.
• Stitching seeds. Generate two versions of a scene and hand them both to a comparison or a batch-aware post-process node in one shot.

The output IMAGE can be split apart again downstream with a batch-split node if you need the individual frames back.

The gotcha

Dimensions must match. image1 and image2 need identical width, height, and channel count, or the node throws a shape error. If your two sources differ - one was upscaled, one wasn't - run them through the same ImageScale first. People hit this constantly and it's almost never the node's fault.

Also worth knowing: this is a pixel-space concatenation. If you want to fuse images in the model's working space instead, that's LatentBlend/LatentComposite (both also wrapped as Icy nodes in this pack) - a different operation entirely.

Installing it

cd ComfyUI/custom_nodes
git clone https://github.com/icekiub-ai/ComfyUI-IcyHider

Restart, or use ComfyUI Manager (search "IcyHider"). No pip dependencies, no models.

Common issues

• Missing node in shared workflows - install the pack or swap in core ImageBatch.
• Shape mismatch errors. Make your two inputs the same resolution first. It's the only real failure mode here.
• "I wanted them side by side." This node batches, it doesn't tile. For a 2-up contact sheet you'd want a composite/concatenate-by-position node, not this one.