ComfyUI Node

ImageShuffle

Randomly reorder four images (and read this before trusting the seed)

By taabata·Created 3 years ago·Updated 2 years ago· 259
ImageShuffle
  • image_1
  • image_2
  • image_3
  • image_4
  • IMAGE
  • IMAGE
  • IMAGE
  • IMAGE
seed0

ImageShuffle is a small utility that takes four images and hands them back in a random order. It exists to support the pack's "mix / shuffle" workflows - the LCM_blender_shuffle.json style graphs where you want to randomize which input lands where on each run, usually to vary which image feeds a blending or generation step. If you're doing repeated batch-style runs and want a different arrangement each time, this is the node.

The inputs

  • image_1 through image_4 - four IMAGE inputs.
  • seed - an INT that looks like it controls the shuffle.

Four IMAGE outputs, in shuffled order.

The gotcha you need to know

Read the source before you rely on the seed: the node calls Python's random.shuffle(newarr) and never actually seeds anything. The seed input is collected and then ignored - the shuffle uses the global random module, which in ComfyUI's Python process behaves like any unseeded global RNG. So:

  • Changing the seed will not reproduce a shuffle. The same seed can give you a different order next run.
  • Re-running the same graph with the same seed does not give you the same output arrangement.

For most uses that's fine - you wanted "surprise me," and you get it. But if your workflow depends on a reproducible order (say, matching shuffled images to a specific seed in a batch), this node will silently betray you. That's a real trap, not a nitpick, because the input's presence implies reproducibility the code doesn't deliver.

Setup

It's part of the taabata/LCM_Inpaint-Outpaint_Comfy pack, so install is the same as the rest:

cd ComfyUI/custom_nodes
git clone https://github.com/taabata/LCM_Inpaint-Outpaint_Comfy
cd LCM_Inpaint-Outpaint_Comfy
pip install -r requirements.txt

Or ComfyUI Manager, search LCM_Inpaint-Outpaint_Comfy. No model downloads needed - this node is pure image plumbing.

Is it worth reaching for?

Honestly, only if you're inside this pack's blender workflows. Any general-purpose random-image-picker or shuffle node in a modern pack does the same job with working seed semantics. This one is a thin, 2023-era utility whose main distinguishing feature is that its seed is decorative. It works for the "just randomize it" case - just don't build anything on top of the seed. If you need deterministic shuffling, you're better off with a node that actually seeds its RNG.

Categoryimage

Inputs (5)

NameTypeDefaultDescription
image_1IMAGE
image_2IMAGE
image_3IMAGE
image_4IMAGE
seedINT00–18446744073709550000

Outputs (4)

NameTypeDescription
IMAGEIMAGE
IMAGEIMAGE
IMAGEIMAGE
IMAGEIMAGE