SenseNova Image List

This is the least glamorous node in the SenseNova-SI pack and the one you'll reach for the moment you want the model to look at more than one image at once. It takes up to ten individual IMAGE inputs and fuses them into a single batched IMAGE output. That's the whole job.

Why it exists

The Generate node's image handling is built around batches: each <image> token in your question corresponds to one image in the batch, so a multi-image prompt ("here's picture A and picture B, which room is bigger?") needs all the images arriving down one wire. ComfyUI doesn't merge separate LoadImage outputs for you - each one outputs its own tensor - so this node is the joining piece. It's pure plumbing in the best sense: no settings, no opinions, just torch.cat(images, dim=0) under the hood. If you've used the utility/aggregation layer of any ComfyUI pack, you know this shape of node.

Inputs and output

Ten optional inputs, unimaginatively named image_1 through image_10, all IMAGE type. All optional, so you can wire up just the two or three you need and ignore the rest. The output is a single images IMAGE batch that plugs straight into the SenseNova SI Generate node's image input.

One thing to know before you lean on it: concatenating along the batch dimension requires every input tensor to share the same height, width, and channels. Feed it a 512×512 and a 1024×1024 and you'll get a tensor-shape error, not a friendly message. If your source images differ in size, normalize them first (an ImageScale or resize node) so they match.

Install

Same shared install as the rest of the pack: clone GACLove/ComfyUI-SenseNova-SI into custom_nodes, clone the upstream OpenSenseNova/SenseNova-SI into deps/SenseNova-SI, pip install -r requirements.txt, restart ComfyUI. There's nothing model-specific about this node - it would work fine in any graph, which makes it a decent generic batch-er if you don't mind the SenseNova name on it.

Gotchas

• Wire nothing and it raises a ValueError ("at least one image input must be connected") rather than passing an empty batch through.
• All images must be the same resolution to concatenate (above).
• Ten is the hard cap. Need eleven? Chain two of these nodes - the first one's output is a normal IMAGE batch, so it can feed the second's image_1 slot alongside fresh images.

That's genuinely all there is. It's the kind of node you'll forget exists until you need to compare three generated images in one prompt, and then it's exactly right.