VNCCS Quad Splitter

One node, two opposite jobs, picked with a mode dropdown: split takes a single image laid out as a 2×2 grid and cuts it into four separate images; compose does the reverse, taking four images and gluing them into one 2×2 grid. The image input and images (list) output stay the same regardless of which mode you're in - what changes is which direction the data flows.

Why you'd want either direction

Generating four related images as one 2×2 grid in a single diffusion pass is a real, commonly used trick - it's cheaper than four separate generations, and for anything where consistency across the four matters (four angles of the same character, four color variants of an outfit) letting the model see all four in one shared context during generation tends to produce more coherent results than generating them independently and hoping they match. compose mode is how you'd build that combined input if you're assembling four existing references into one grid to feed a model. split mode is what you run afterward to pull the model's 2×2 output back apart into four usable individual images.

Inside VNCCS specifically, this kind of batched-grid generation is consistent with how the pack handles previews - Character Creator's "Generate Preview" button, per the README, lets you see what a character will look like without running the full heavy generation, which is exactly the kind of thing a cheap grid pass followed by a split is good for.

The one input that matters

Just mode - split or compose. There's no crop or alignment control here; it's a literal geometric quarter of whatever image you feed it in split mode, and a straightforward 2×2 tiling of whatever four images you feed it in compose mode. If your source images aren't the same dimensions in compose mode, expect the resulting grid to have visibly mismatched panel sizes - this node isn't doing any resizing for you.

How to install it

Ships with the main pack:

• ComfyUI Manager - search VNCCS - Visual Novel Character Creation Suite, install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/AHEKOT/ComfyUI_VNCCS.git, then cd ComfyUI_VNCCS_Utils && pip install -r requirements.txt, restart, and let Comfy Manager grab any missing dependencies.

Common issues

Split output panels have visible seams or slightly different content than expected. Check that the source image is actually a clean, evenly-divided 2×2 grid - an odd aspect ratio, or a grid generated with visible padding/borders between panels, will split at exactly the geometric quarter regardless of where the actual panel boundaries are, cutting into content rather than along the seam.

Compose mode produces a grid with mismatched panel sizes. As noted above, this node doesn't resize inputs to match - run your four images through VNCCS Resize to a common size first if they came from different sources or generations.

Confusing this with VNCCS Sheet Extractor. If you're working with VNCCS's larger sprite sheets (the 2×6, twelve-panel layout used elsewhere in the pack) rather than a simple 2×2 grid, you want VNCCS Sheet Extractor instead - this node is specifically 2×2 and has no concept of indexing into a bigger grid.