GR Flip Tile Random Inverted

A mosaic/glitch-art effect node: it slices an image into a grid and randomly flips or color-inverts a subset of the tiles, giving you the kind of broken-mirror, corrupted-photo look that's genuinely hard to fake by hand tile-by-tile. This is a creative-effects node rather than anything with a KB-documented technique behind it - the mechanism is straightforward enough to explain from the schema and category name alone: it belongs to this pack's Tiles category alongside its sibling GR Flip Tile Random Red Ring, which does the same grid-flip idea with a colored ring highlight instead.

How it works

columns and rows set the grid the image gets sliced into. Some fraction of those tiles get randomly flipped and/or color-inverted rather than staying untouched - the randomness isn't seeded on this particular node (unlike its Red Ring sibling, which does expose a seed), so expect a different result each run unless you're also controlling upstream randomness. border and colour (seven choices: black, white, red, blue, green, purple, yellow) add a dividing line between tiles.

The inputs and outputs that matter

• image - the source to tile and randomly invert.
• columns / rows - grid density. More tiles means a finer-grained, busier effect; fewer means larger, more obvious inverted blocks.
• border / colour - optional dividing line between tiles, off by default (border defaults to 0).

Two outputs, both named IMAGE in the schema with no further distinction between them - the schema doesn't label what differs between the two, so the practical approach is to preview both and see which one is the result you want for your case rather than assuming based on port order.

How to install it

ComfyUI Manager, search GraftingRayman. Manual install:

cd ComfyUI/custom_nodes
git clone https://github.com/GraftingRayman/ComfyUI_GraftingRayman

Restart. As with every node in this pack, OpenAI's CLIP needs installing separately or nothing imports - portable ComfyUI: .\python_embeded\python.exe -m pip install git+https://github.com/openai/CLIP.git; system Python: pip install git+https://github.com/openai/CLIP.git.

Common issues & troubleshooting

Pack fails to load. Check your startup console for the CLIP import error and run the pip install above.

Result looks the same every time despite wanting variation. This node doesn't expose a seed input, so if you need a reproducible result, you can't get one directly from this node - that's what its Red Ring sibling is for, which does take a seed.

Effect is barely visible. With a coarse grid (low columns/rows) and only a handful of tiles randomly selected, the effect can be subtle. Increase tile density for a busier, more obviously glitched result.

Not sure which of the two IMAGE outputs to use. Preview both - the schema doesn't distinguish them by name, so there's no shortcut here beyond checking your actual output.