ExtendImageList

ExtendImageList concatenates multiple IMAGE lists into one ordered list. Where its sibling Extend nodes work on numbers and strings, this one deals in pixels - which is where list concatenation gets genuinely useful and surprisingly easy to get wrong.

Why you'd reach for it

Images arrive in lists from three common places: ImageBatchToList (also in this pack), batch outputs from samplers, and video frame nodes. When you've got two batches you want to process as one - say, half your frames come from one pass and half from another, or you're assembling a comparison grid from several generations - ExtendImageList is the merge point.

The grid workflow is the standout. Grid nodes want a single image list to tile; with ExtendImageList you can assemble the tiles from multiple sources in a chosen order instead of hoping each source happens to emit everything you need in one shot. For a row-by-row comparison grid, that control over ordering is the whole trick.

How it works

The inputs_len widget (default 2, up to 32) sets how many image-list inputs exist. As with the rest of this family, changing it does nothing until you right-click → "update Inputs" (the README's ❔ marker). The node then appends each input list in order.

Each socket is typed as a list of images. Watch out: a single IMAGE from most samplers is already a batch of frames, and the distinction between "one image" and "a list of images" is where beginners trip. If a connection won't snap in, the upstream node probably outputs a plain image or a batch, not an image list - ImageBatchToList converts a batch into a proper list.

Install

Same pack, one install:

cd ComfyUI/custom_nodes
git clone https://github.com/bmad4ever/comfyui_bmad_nodes

then restart ComfyUI. Manager users: search "comfyui_bmad_nodes". The pack's Python dependencies get installed alongside; this node itself is pure list handling.

Common issues

• Inputs not appearing. Change inputs_len, right-click, "update Inputs".
• Order is everything. Left-to-right, top-to-bottom, so arrange sources deliberately - especially for grids, where a swapped input scrambles the whole layout.
• Batch vs list confusion. A sampler's IMAGE output is a tensor batch, not a list. Run it through ImageBatchToList first if your merge partner is a list-based node.

It's a dumb, dependable connector - and for grid work, dumb and dependable beats clever every time.