ComfyUI Node

Split Images

Split an image batch into ten named sockets and stop fighting the list

By yolain·Created 3 months ago·Updated about 17 hours ago· 158
Split Images
  • images
  • IMAGE0
  • IMAGE1
  • IMAGE2
  • IMAGE3
  • IMAGE4
  • IMAGE5
  • IMAGE6
  • IMAGE7
  • IMAGE8
  • IMAGE9

Every now and then ComfyUI gives you an IMAGE batch - say, a directory loader returns ten frames - and the node you want to feed wants exactly one image. Split Images is the shortcut: one input, ten fixed outputs (IMAGE0 through IMAGE9), each a single image. It accepts either an image list or a single batched tensor, and it unwraps them into individual items you can wire wherever you need a discrete image.

Why this exists in a video pack

In ComfyUI-Easy-Media's world, images come out of the timeline and multitrack editors as batches - reference frames, keyframes, guide images. The splitter is what turns "the reference sheet for this segment" into "reference number one, reference number two" so you can hand each to a different conditioning node, a LoRA, or a first/last-frame setup. It's the same job easy makeImageList does in reverse, and the two pair naturally: make a list, process it, split it back.

It's also a genuinely handy general-purpose utility outside the pack's own workflows - if you've ever wished an image batch came with named sockets, this is that.

Behavior worth knowing

The split is positional: IMAGE0 is the first image in the batch, IMAGE1 the second, up to ten. Empty slots come out as None, which lets you leave outputs dangling safely. Two practical limits to keep in mind:

  • The ten-output ceiling is fixed. More than ten images in and the extras silently don't get an output - trim your batch first or use a list-indexing approach.
  • Batches and lists are both handled, but a list of lists isn't - keep it to one level deep.

If your pipeline is "load ten reference frames, condition each one separately," this node is the exact shape you want. If you're processing hundreds of images, this is the wrong tool; you want a loop with dynamic indexing instead.

Install

It's part of ComfyUI-Easy-Media:

cd ComfyUI/custom_nodes
git clone https://github.com/yolain/ComfyUI-Easy-Media.git

restart ComfyUI (ComfyUI Manager, "Easy-Media", also works). No model files needed - this one's pure tensor wrangling, so there's nothing extra to download. FFmpeg being present matters for the pack's video side, but this node won't touch it.

The trap

The fixed ten sockets. People routinely load a batch of 12, wire up IMAGE0IMAGE9, and wonder where the last two frames went. There's no error - ComfyUI just never sees them. If your batch size varies, count before you split. Otherwise this is about as frictionless as ComfyUI gets: plug in, pull out single images, move on with your life.

CategoryEasyUse/Image

Inputs (1)

NameTypeDefaultDescription
imagesIMAGE

Outputs (10)

NameTypeDescription
IMAGE0IMAGE
IMAGE1IMAGE
IMAGE2IMAGE
IMAGE3IMAGE
IMAGE4IMAGE
IMAGE5IMAGE
IMAGE6IMAGE
IMAGE7IMAGE
IMAGE8IMAGE
IMAGE9IMAGE