Split Image ♾️Mixlab
Cut an image into a grid of regions
- image
- grids
- grid
- mask
Split Image carves an image into a grid of cells and hands you back that grid as a structured thing you can address region by region. On its own that sounds like a tiling utility, but the reason it exists in this pack is the node it feeds: Mixlab's advanced GLIGEN apply, which places different text prompts into different grid cells. Split Image defines the layout; GLIGEN fills each cell with its own concept. Together they're a regional-prompting / spatial-layout tool - "put a cat here, a house there" by dividing the canvas.
So think of this less as "chop an image into thumbnails" and more as "define the spatial grid that a downstream node will assign prompts to."
How it works
You give it an image and a cell count, and it produces grid data - the regions the image is divided into - plus a matching mask. The output isn't a plain batch of image crops; it's a Mixlab _GRID type that carries the layout structure, which is exactly what GLIGENTextBoxApply_Advanced expects on its grids input. That coupling is the whole design: Split Image is the layout half, GLIGEN is the prompt-placement half.
Inputs and outputs
image(IMAGE, required) - the source image whose dimensions define the grid canvas.num(INT, default4) - how many cells to split into. This is your layout dial - more cells, finer spatial control.seed(INT, default4) - a seed that influences how the split is arranged, so you can reroll the layout.
Outputs:
grids(_GRID) - the collection of regions, plural. This is what you feed into GLIGEN's advanced apply.grid(_GRID) - a single grid representation.mask(MASK) - a mask matching the split, useful if you want to operate on regions with mask-based nodes.
The _GRID outputs are Mixlab-specific, so they're meant for the pack's own grid-aware nodes rather than generic image nodes.
How to install it
Ships with the Mixlab pack.
- ComfyUI Manager: search
comfyui-mixlab-nodes, install, restart. - Manual:
cd ComfyUI/custom_nodes && git clone https://github.com/shadowcz007/comfyui-mixlab-nodes.git, install requirements (install.baton Windows or../../../python_embeded/python.exe -s -m pip install -r requirements.txt), restart.
Common issues
The _GRID output won't connect to my node: it's a Mixlab custom type, so it only plugs into nodes that accept _GRID - chiefly GLIGENTextBoxApply_Advanced in this same pack. If you were hoping to feed it into a generic image-batch or tiling node from another pack, that won't link; use the mask output or extract image crops another way if you need plain images.
I just want tiles, not a GLIGEN layout: then this is probably the wrong node - it's built around the grid-prompting pipeline. For straightforward tiling/splitting into separate images, a general image-grid or crop node will be less surprising.
Layout isn't what I expected: change num for a different cell count, and reroll seed to get a different arrangement. Remember the grid is defined against the input image's aspect ratio, so a very wide or tall image will grid differently than a square one.
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | — | |
| num | INT | 41–500 | — |
| seed | INT | 41–500 | — |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| grids | _GRID | — |
| grid | _GRID | — |
| mask | MASK | — |