Pixelate / Mosaic Image
Create a pixelated mosaic version of an uploaded image.
Overview
The Pixelate / Mosaic tool replaces an uploaded image with a grid of larger, blocky cells, each filled with the average colour of the source region it covers. Choose a cell size — 8, 16, 32 pixels or more — and the tool returns a re-encoded copy with the mosaic effect baked in.
It is the standard way to redact a face or licence plate in a screenshot, to give an image a retro 8-bit feel for a social card, or to generate an abstract preview from a hero photo that's still recognisable at thumbnail size. Compliance and privacy teams use this exact technique to redact sensitive parts of an image online before publication.
How it works
The image is partitioned into a grid where each cell is the chosen pixel size square (typically 8, 16 or 32). For each cell the tool computes the mean colour of every input pixel inside it — (sum R) / count, (sum G) / count, (sum B) / count — and fills the entire cell with that single colour. The output therefore has the same dimensions as the input but only one distinct colour per cell.
For variable-size pixelation, advanced workflows can run k-means on the source region to pick more representative colours, but a plain average is the standard and is what users expect when they see "mosaic". The output is encoded losslessly in PNG by default so the hard cell boundaries stay crisp.
Examples
Before: portrait.jpg (4000x6000)
After: portrait-mosaic.jpg with 32 px cells — face still recognisable
but unmistakably stylised.
Before: licence-plate-photo.jpg
After: cells over the plate region only (after a crop pass) — text
unreadable, sensitive info redacted.
Before: hero-banner.jpg
After: hero-mosaic.jpg with 16 px cells — useful as an artistic
loading placeholder.
FAQ
What cell size redacts a face?
Choose a cell size of roughly the eye-to-eye distance in pixels. For a 200 px-wide face, 16-24 px cells make features unrecognisable while keeping the image visually balanced.
Is this the same as a Gaussian blur?
No. Blur creates a smooth, continuous reduction in detail. Pixelation creates hard, blocky cells. For redaction, both work; legally, pixelation is sometimes preferred because the original information cannot be recovered by sharpening.
Can pixelation be reversed?
A single average per cell discards almost all the information in that region. Reversal is essentially impossible without external data — far safer than mild blur for sensitive redaction.
Does it preserve transparency?
Yes — the alpha channel is averaged per cell along with RGB, so transparent regions stay transparent.
What's the largest cell size that's useful?
Beyond ~64 px cells the image becomes near-unrecognisable. For artistic effects, 8-32 px is the sweet spot.