DD Image Splitter

DD Image Splitter cuts an image into strips and hands you back exactly one of them. Left-right or top-bottom, 2 to 10 pieces, with custom ratios like 2:1:3 instead of always-equal slices. It's a small node, but it solves the specific awkward problem of "I have a wide image and I need the middle panel as its own image."

How it works

Three settings drive the cut: 切分方向 (Split Direction - left/right or top/bottom), 切分份数 (Number of Parts, 2–10), and 切分比例 (Split Ratio), a string like 1:1 or 2:1:3. The ratio string is the interesting part - the number of values must match the part count, and they're normalized, so 2:1:3 across a 1200px-wide image becomes 400px / 200px / 600px. If the ratio count doesn't match the part count, or the string is garbage, it silently falls back to equal slices - worth knowing, because a typo like 1:1:2 on a 2-part split just quietly gives you 50/50 instead of erroring.

输出位置 (Output Position) picks which strip survives. Position is 1-based (first slice = 1), and if you ask for a position beyond the part count it clamps to the last slice. The single 图像 (Image) output is just the one selected strip - the others are discarded, not emitted. So if you want all four panels of a comic strip, you need four of these nodes or a different tool; this node's whole design is "extract one region."

When you'd reach for it

Comic and storyboard pipelines are the obvious home: cut a generated multi-panel image into individual panels for upscaling or inpainting one at a time. It's also handy for cropping letterboxed or split-reference images (like a before/after grid you want to separate), and for grabbing a specific slice out of a wide composition without manual x/y math. Because the output is a normal IMAGE, it feeds straight into a sampler, an upscaler, or a VAE encode.

Installation

Standard DD-Nodes install:

cd ComfyUI/custom_nodes
git clone https://github.com/Dontdrunk/ComfyUI-DD-Nodes
cd ComfyUI-DD-Nodes && pip install -r requirements.txt

Restart ComfyUI or use ComfyUI Manager (search "ComfyUI-DD-Nodes"). No extra models. The node works through PIL under the hood, which the pack already depends on.

Honest take

The "one strip out" design is both the strength and the limitation. If you genuinely need a single slice - which, honestly, is most of the time people split images - it's more direct than a crop node because the ratio math is handled for you. If you need all the parts, this is the wrong node and you'd be happier with a batch splitter. Also note it processes one image at a time from the batch (it takes the first frame of the input), so don't feed it a video and expect per-frame output. Like the rest of the pack, labels are Chinese-first with an English locale available in ComfyUI settings.