Plenty of image output arrives as a grid — a 2×2 set of variations, a four-up contact sheet, a supplier's product sheet with nine items on one JPEG. As a single file it's useless as ad creative: you can't post a tile, run one through an upscaler, or feed one into an image-to-video generation. split_image_grid takes the image plus a row and column count and returns the individual cells.
It's a deterministic slice, not a model call — you're telling it the grid is 2 by 2 or 3 by 3, and it cuts accordingly. That makes it reliable on a real grid and wrong on an image that only looks like one.
Powered by
Split an image into a grid (NxM) of smaller images. Useful when a user provides a grid image and wants to extract individual tiles or cells from it.
What to tell the agent
Versely's agent maps this plain-English request directly onto split_image_grid. You don't need to know the parameter names — just describe what you want.
“This is a 2x2 grid of variations — split it into four separate images.”
How it works
1. Count the grid
rows and cols are both required, alongside image_url. Get them right: a 2×2 instruction on a 3×3 sheet slices straight through the middle of six tiles.
2. Run the split
The tool cuts the image into rows × cols cells and returns each as its own image.
3. Pick the ones worth keeping
Usually one or two of a variation grid are the keepers. The rest can be discarded now rather than carried through the rest of the pipeline.
4. Take the keeper further
A single tile is a quarter of the original's pixels, so upscale_image is the usual next step before print or a large placement — and the tile is now a valid input for generate_image_from_image or an image-to-video generation.
What it costs
A deterministic slice rather than a generation, so it's a lightweight image operation. The cost that follows it is usually the upscale — that's billed per the model you pick, so confirm with estimate_cost before running it across every tile.
The formula behind that number: What does a batch of 100 AI product images cost? — charges one fixed price per call, so the image's size and shape are free.
Limits & things to know
- It cuts on an even grid at the row and column counts you supply — it does not detect the grid, and it won't handle an irregular collage or tiles separated by uneven margins.
- Each tile is a fraction of the source resolution. A 1024px 2×2 grid yields four 512px images, which is often too small to publish without an upscale.
- Borders, gutters and any caption strip on the original get sliced along with the picture — trim or upscale afterward if the source had visible dividers.
Who uses this
- Pulling the one good variation out of a 2×2 generation grid
- Breaking a supplier's multi-product sheet into separate product images
- Turning a contact sheet of frames into individually usable stills
- Extracting tiles to use as reference images for follow-up generations
Frequently asked questions
Does it detect the grid automatically?+
No — rows and cols are both required parameters. You tell it the shape of the grid; it cuts evenly on those lines.
Will the split tiles be full resolution?+
Each tile is its share of the original's pixels, so a 2×2 split gives you quarter-size images. Pair it with upscale-image-resolution if the tile needs to be published larger.
Can I use a split tile to make a video?+
Yes — once separated, each tile is a normal image URL, so it works as an input to generate_image_from_image or an image-to-video generation the same as any other picture.
Related Versely tools
Related editing jobs
Split a Grid Image Into Separate Images inside Versely
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync in a single app.