Nodes/comfyUI_LLM/📂 创建空图像Batch
ComfyUI Node

📂 创建空图像Batch

An 'empty' image batch exists, and this node makes one

By XieJunchen·Created about a year ago·Updated about a month ago· 2
📂 创建空图像Batch
  • images
  • image_batch

An empty image batch sounds like a contradiction - but in ComfyUI an IMAGE is a tensor with a batch dimension, and "zero images" is a perfectly valid (if useless) tensor. This node, "📂 创建空图像Batch", is the seed for a pattern the whole pack is built around: start from nothing, then append frames one at a time as a loop or pipeline feeds them in. It's a utility, and a useful one, once you've met its partner node.

How it works

Called with nothing connected, it returns a zero-size tensor - shape (0, 0, 0, 0), no actual pixels. That tensor is the "empty batch" contract that AppendImagesToBatch knows how to consume: append to it and you just get the added images back, which is how you bootstrap a growing stack.

There's a second mode hiding in here. Wire an images input in and it stops being an "empty" node entirely - it becomes a channel normalizer. It strips the alpha channel from RGBA images and repeats grayscale to three channels, then returns the batch (and if it gets a single 3D image it promotes it to a batch of one). So it doubles as "make sure my batch is a clean 3-channel 4D tensor."

Inputs and output

  • images (IMAGE) - optional. Leave it empty for the zero-size batch, or connect it for the pass-through-and-normalize behavior.
  • Output image_batch (IMAGE) - either the empty tensor or the cleaned-up input batch.

The channel thing matters more than it looks

That to_3ch logic exists because this pack generates 4-channel RGBA in places - LoadGifFromLocal outputs RGBA frames, for one - and many downstream nodes expect plain RGB. If you've got an RGBA image and a stubborn node complaining about channels, routing it through this one is a legitimate fix.

Installing it

Standard pack install - ComfyUI Manager → search comfyUI_LLM, or:

cd ComfyUI/custom_nodes
git clone https://github.com/XieJunchen/comfyUI_LLM
# restart ComfyUI

Pure torch, no extra dependencies.

Honest assessment

On its own this node is a zero. Its whole value is as the image_batch seed for AppendImagesToBatch, or as a channel-normalizing utility. If you're not building batch-assembly loops, you'll probably never use it - and that's fine. ComfyUI's own ecosystem has batch-combining nodes that do the cat job, but this pack's two nodes are designed to work together, and together they give you a clean way to accumulate frames in the exact order you feed them. If you want to build a batch by hand, this is where that hand starts.

Category云服务/集合工具

Inputs (1)

NameTypeDefaultDescription
imagesoptIMAGE

Outputs (1)

NameTypeDescription
image_batchIMAGE