ZML_遮罩分割-五

The "more strips" version of ZML_MaskSplitNode. Same idea - no mask input, pure geometry, you give it a canvas size and it manufactures the splits - but instead of two halves you get up to five equal vertical strips.

If you've ever built a workflow that needs "region 1, 2, 3, 4, 5 across the image, each with its own prompt or its own processing," you know the annoying part is usually just generating those five masks. This node is that annoyance removed: it always emits five full-size masks, and the 分割数量 slider decides how many of them actually contain a white strip.

The inputs and outputs

• 宽度 / 高度 (default 1024 each) - the canvas. Match to your generation resolution; nothing here reads your actual image.
• 分割数量 (1–5, default 5) - how many strips to actually produce. Set to 3 and you get three filled strips; masks 4 and 5 come out empty (all black).

Outputs:

• 遮罩1 through 遮罩5 - each a full-size mask. For 分割数量 = N, strips 1..N are white bands of equal width and the rest are empty.
• 单独遮罩 - a single cropped strip at minimal width (the width of one segment), useful when you want one representative region as a small tensor.
• 完整遮罩 - an all-white full-canvas mask.

The strips are always vertical, by the way. Unlike the two-way ZML_MaskSplitNode, there's no 横/diagonal option here - if you need horizontal or diagonal splits, use the two-way node instead.

What it's for

The obvious use is regional conditioning across a wide canvas: five columns, five prompts, one composite. It also slots into the ZML pack's sub-workflow and list machinery - since each mask is just a tensor, you can treat 遮罩1..5 as five values to fan out to five copies of a sub-pipeline. Given the pack's recent push on lists and batching (the author added "批次转列表" and list-output nodes around the same era), thinking of these as a poor-man's batch is the right mental model.

The quirk to know

单独遮罩 is only ever the first segment. It's not the currently-active strip or a crop of the whole thing - the code hard-codes it to the width of segment one. Fine if that's all you need; mildly surprising if you assumed it tracked 分割数量. It's a small, harmless detail and the kind of thing that tells you this node is built for one specific workflow rather than generality.

Install

Standard pack install - this is one node of 160+ in ComfyUI-ZML-Image:

cd ComfyUI/custom_nodes
git clone https://github.com/zml-w/ComfyUI-ZML-Image
# restart ComfyUI

or via ComfyUI Manager (search "ComfyUI-ZML-Image"). No dependencies beyond the pack baseline. Chinese-first UI; English users, get the translation patch at https://github.com/zml-w/ZZZ_ZML_English_Patch.

One-person project, so treat everything with the author's own caveat: nodes he uses constantly are polished, others may have rough edges. This one is dead-simple numpy masking, so it's been reliable - the only thing to trip on is the 单独遮罩 behavior above.