Extract a Color Palette From an Image

No. The image is decoded, scaled down, and analyzed entirely in your browser using a canvas. No bytes are sent to a server, so private screenshots, artwork, and photos stay on your device.

The tool uses median-cut quantization. It samples the pixels, repeatedly splits the color box with the widest channel spread at the midpoint of that channel, then averages each final box into one representative color. Swatches are ranked by how many sampled pixels fall into each group.

The tool only splits groups that still contain a spread of colors. If an image has just a handful of distinct colors, such as a flat logo or a screenshot with solid fills, it returns those real colors instead of inventing extra near-duplicates.

Yes. The image is scaled so its longest edge is about 200 pixels before sampling, which keeps the analysis fast on large photos. The palette reflects the whole image because scaling blends nearby pixels rather than dropping regions.

No. Pixels below an alpha threshold are dropped before analysis, so a transparent PNG background does not pollute the palette with a phantom color. Only reasonably opaque pixels contribute to the swatches.

Not quite. Each swatch is the average of a cluster of similar pixels, so it represents a region of the image rather than one exact pixel. For an exact single-pixel value, use the image color picker instead.

It is the share of sampled pixels assigned to that color group. The colors are sorted from the largest share to the smallest, and the proportional bar at the top shows the same weighting as a single stacked strip.

Click a swatch to copy its hex code on its own. Copy hex list gives a comma-separated list of every hex in the palette, and Copy CSS wraps the palette in a :root block with numbered custom properties ready to drop into a stylesheet.

Yes. The algorithm is fully deterministic with no random seeding, so re-running it on the same image at the same color count produces an identical palette every time.