This pulls a palette of several distinct colors out of an image, rather than a single swatch - useful when you need to know the two, four, or eight colors that actually make up a design, not just one blended average. Processing happens entirely on a canvas in the browser; the image is never sent anywhere.

How the palette is chosen

The image is sampled at a fixed 80×80 resolution and every pixel's color is grouped into buckets (each channel rounded to the nearest step of 32), then the buckets are ranked by how many pixels fall into them - the same frequency approach the average/dominant tool uses, but extended to keep several top candidates instead of just the single winner.

To avoid returning a palette of near-identical shades - several slightly different blues, say - each candidate is only added if it's at least a minimum RGB distance away from every color already picked. That's why the result tends to look like genuinely distinct colors rather than gradations of the same one.

The slider controls how many colors to extract, from 4 to 12. Raising it doesn't necessarily fill every slot - if the image doesn't have enough sufficiently distinct colors, the extracted list will simply be shorter.

When you'd use this

Reverse-engineering a color palette from a photo, mockup, or piece of artwork to reuse in a design system, or checking which colors visually dominate a product photo before writing marketing copy around it.

Frequently asked questions