Pull the dominant colours out of a photo and copy them as HEX, RGB, HSL, CSS variables or JSON, with the share of pixels each one covers.
Colour sampling happens in this tab. Nothing is uploaded.
The dominant colours and their HEX, RGB and HSL codes appear here.
Swatches with HEX, RGB and HSL codes appear here.
Shares are measured over opaque pixels only. Transparent areas contribute no colour, and very similar shades are merged so the list names colours a person would actually use.
Your file is never sent to a server. It is read and processed by your own browser, and the results are not stored.
The image is drawn onto a small canvas, its opaque pixels are bucketed by colour, and the buckets are ranked by pixel count. Nearby buckets are merged so the list does not repeat almost-identical shades. The percentage beside each swatch is the measured share of opaque pixels.
This is deliberately not a k-means or median-cut library call. A deterministic bucket-and-merge pass is fast enough to run on every slider change and produces the same answer for the same image.
A photograph’s dominant colours are averages, not brand decisions. They are a useful starting point for a mood board or a background tint, but a design system needs contrast checks and a limited type scale, neither of which this tool claims to produce.
The CSS export uses numbered custom properties so the palette can be dropped into a theme file. Keep the comments: they record the measured share, which is the part a future reader cannot recover from the hex value alone.
No. Image Color Palette Extractor runs in your browser, and the file is read there rather than sent to a server for processing. Our host still sees ordinary page-request metadata, which is described on the privacy page.
Yes, Image Color Palette Extractor is free, with no paid tier, no account, and no watermark added to your output. The limits are technical rather than commercial: file size, page count and supported formats are listed on this page.
Image Color Palette Extractor supports .jpg, .jpeg, .png, .webp files up to 25MB directly in your browser.
Opaque pixels are sampled at a reduced size, grouped into colour buckets, and ranked by how many pixels each bucket covers. Very similar shades are merged so the result names colours a person would use rather than a dozen imperceptibly different blues.
Transparency has no colour to measure. Fully transparent pixels are skipped, and partly transparent pixels contribute their visible colour. The displayed share is therefore a share of opaque pixels, not of the whole canvas.
Yes. The CSS export writes custom properties with numbered names and a comment showing each share, and the JSON export includes HEX, RGB, HSL and the percentage. Neither is a full contrast-checked palette, so check text contrast separately.