🍒Image_List_Splitter📂图片列表分割器

Somewhere in every batch-heavy ComfyUI workflow you hit the same wall: a node upstream produced a stack of N images, and the node downstream only wants M of them - or you want to process M and keep the rest aside for later. Image_List_Splitter is the pack's answer: take a list of images, cut it into "the first (or randomly-chosen) N" and "everything that's left," and hand you both lists.

It's a pure plumbing node, the kind you install and forget until you actually need it, at which point it saves you from writing a custom list-slicing node or fighting with a batch-to-list converter. It's in the pack's 图像处理 (image processing) section.

How it works

The node is list-aware on both ends (INPUT_IS_LIST and OUTPUT_IS_LIST are both on). Give it an images list, and it splits by split_count - either from the front (fixed order) or, with enable_random on, by shuffling with random_seed first. The first group goes to selected_images, the remainder to remaining_images.

Two edge behaviors worth knowing. First, split_count is clamped to the actual list length, so asking for 50 from a 10-image list just gives you all 10. Second, if the remainder is empty, the node doesn't return an empty list - it returns a single black placeholder image. That's deliberate (downstream nodes often choke on empty lists) but it means "nothing left" looks like "one black image left," so don't be surprised by a black frame at the end of your pipeline.

The inputs that matter

• images - the image list to split.
• split_count - how many go into the first group.
• enable_random - off = first N in order; on = random N.
• random_seed - only used when random is on; gives you a repeatable "random" pick.

Outputs: selected_images (list) and remaining_images (list).

Installation

Standard KimNodes pack install - Manager → search "ComfyUI_KimNodes" → Install → Restart, or git clone https://github.com/wjl0313/ComfyUI_KimNodes into custom_nodes. No extra dependencies.

Common issues

• The node behaves oddly with a single image - it's designed around lists. If you feed a non-list IMAGE, it wraps it into a one-element list, so "split 1" just returns the same image in both outputs' placeholders. Make sure upstream is actually a list.
• Black image at the end - that's the empty-remainder placeholder, not corruption. If you genuinely need "nothing" rather than "black," put a filter node downstream.
• Random selection not random - you set the seed once; same seed, same "random" pick. Set random_seed fresh or vary it if you want variety.

It's not glamorous, but for batch splitting it's a clean, dependency-free tool. And unlike a few nodes in this pack, it does exactly what its description says.