Image to palette
Clustered in OKLab and split by luminance. No model, no guessing — the same frame always gives the same palette.
Processed on your device. Your image never leaves your browser.
Drop a photograph or film still
Drop a frame and its dominant, highlight, midtone and shadow colours appear here, with the warm–cool balance and average chroma.
Extract a colour palette from any image
Drop a photograph or a film still and the tool reads its dominant colours, then splits the frame by luminance into highlights, midtones and shadows and reports the most saturated colours separately. It also measures the frame: how much of it is warm against cool, its average chroma, and its dominant hue.
The clustering happens in OKLab rather than RGB, because a clusterer minimises distance and distance in RGB is not perceptual — it merges two greens a person sees as distinct and splits a gradient a person reads as one colour. It is deterministic: the same frame gives the same palette every time, and every colour in the result can be explained.
The image is decoded and sampled in your own browser and handed to a worker in the same tab. It is never uploaded. For anyone holding unreleased client footage, that is the whole reason to use this rather than an online extractor.
How to use it
- Drop an image onto the frame, or choose a file. JPEG, PNG, WebP and AVIF stills all work.
- Read the dominant palette, then the highlight, midtone and shadow bands and the most saturated colours.
- Check the warm–cool balance, average chroma and dominant hue for the frame as a whole.
- Open the dominant colours in the palette generator to keep working with them.
Questions
- Is my image uploaded anywhere?
- No. The file is read into a canvas in your browser, downsampled, and clustered in a Web Worker on your own machine. Nothing leaves the tab.
- Why do the highlight and shadow bands each cover a third of the frame?
- The bands are split by rank, not by fixed thresholds. A low-key frame has no pixels that are bright in absolute terms, but it still has a brightest third, and that is what the highlights should show.
- Is this AI?
- No. It is k-means clustering in OKLab with luminance segmentation — arithmetic, not a model. The same image always gives the same answer, and there is nothing to guess about why a colour is there.
More free colour tools
- GenerateBuild a palette on real colour-theory relationships, in OKLCH.
- InspectEvery value for one colour, with its tints, shades and harmonies.
- ContrastWCAG contrast, with the nearest accessible colour when it fails.
- ScalesPerceptually even 50–950 ramps for a design system.
All of Color Lab runs in your browser. Nothing is uploaded and no account is needed.