HappyHorse 多路图像批量合并

Not every node in an API pack calls an API. HappyHorse ImageBatchMulti ("HappyHorse 多路图像批量合并") is this pack's quiet utility: it merges up to 9 separate IMAGE inputs into a single batch tensor, and when those images are different sizes, it resizes them all to image1's dimensions so the batch stays rectangular. No DashScope key, no OSS, no cost - it's plain local torch.

Why you'd reach for it

Its natural job in this pack is feeding HappyHorse R2V, whose reference_images input expects one IMAGE batch (1 to 9 frames). ComfyUI's stock ImageBatch node does the merging but chokes on mismatched dimensions; this node handles the resize for you, bilinear-interpolating everything to image1's size. So a realistic R2V prep chain is:

LoadImage → HappyHorse ImageBatchMulti → HappyHorse R2V (reference_images)

with each reference image getting its own LoadImage into one of the nine input slots.

Inputs and output

• image1 (required) - sets the output dimensions; everything gets fit to it.
• image2 through image9 (optional) - wire any subset you need.
• images (output) - one IMAGE tensor whose batch size equals however many inputs you connected.

It's not an output node, so it needs a downstream consumer - an R2V node or any batch-capable node. Two details worth knowing: the 1–9 limit matches R2V's reference cap exactly, so you can fill all nine slots and feed them straight through; and it's a single-frame merge per input, not an animation-time thing.

Install and one honest take

The pack install covers it: ComfyUI Manager (search "Comfyui-Wenyi-demo1") or git clone https://github.com/WY-S/Comfyui-Wenyi-demo1 into custom_nodes plus pip install -r requirements.txt, then restart. Nothing extra - this node pulls in no dependencies beyond what the pack already needs.

If your source images are already the same resolution, stock ImageBatch does this job and you don't need this node. It earns its keep exactly when you're mixing resolutions and don't want to add a ResizeImage per input just to feed R2V. In that specific corner it's the least fiddly thing in the pack.