Color planning
You have a photo of a painted miniature, a color scheme card, a piece of artwork or a poster, and you want the colors out of it in a form you can paint from. Drop it in and the page reads the pixels on a local canvas and hands back the dominant colors as hex codes, with how much of the picture each one covers and a rough hue family to hunt for on your rack.
This reads pixel colors only. It does not identify paints and does not match hex codes to pots from any brand.
A model on a plain background gives the cleanest read: daylight, no flash, and the figure filling most of the frame. A phone shot on a blue cutting mat works, but tick the backdrop trim so the mat does not become your palette.
The photo is drawn onto a canvas no bigger than 480 pixels on its longest edge, then sampled on an even grid. That is why a 12-megapixel photo is not slower than a small one, and why nothing has to be sent anywhere to be counted.
The page splits the sampled colors into boxes, always halving the box with the widest spread of color, and repeats that until it has as many boxes as you asked for. Each swatch is then a real pixel from your photo — the one sitting closest to its box's average, not a blend — and its share is counted by assigning every sampled pixel to whichever swatch is nearest. That is what makes the percentage a coverage figure instead of a guess.
The hue families tell you which part of your rack to look at, and the starting plan turns the list into a base, a shade, a highlight and a spot color. From there it is your eyes and your own paint — this page never guesses a pot number.
Everything below was calculated by the same functions the page runs in your browser, from the built-in sample photo — a small painted figure in steel armor and a crimson cloak, photographed on a blue cutting mat. It is the palette the tool shows before you load anything of your own, so you can check the arithmetic against what is on screen.
| Swatch | Coverage | Family | Value |
|---|---|---|---|
| #8A8D9A | 26.3% | Neutrals, grays & blacks | mid-tone |
| #68472D | 18.9% | Browns & tans | shade |
| #8D9296 | 16.6% | Neutrals, grays & blacks | mid-tone |
| #684B32 | 16.2% | Browns & tans | shade |
| #4C5159 | 12.9% | Neutrals, grays & blacks | shade |
| #8E2D35 | 9.0% | Reds | shade |
`Read it like a photo, not a paint list. The biggest swatch came back at $#8A8D9A with $26.3% of the pixels, and the smallest at $#8E2D35 with $9.0%. Before you decide to buy a pot for a color, look at that share: it is how much of the model that color actually has to cover.`
`The crimson cloak is the useful lesson here. It is the color your eye goes to, and it still only owns $9.0% of the pixels — $#8E2D35 — because a cloak is a small part of a figure standing on a big mat. A photo palette rewards whatever covers the most area, so pick your spot colors on purpose instead of by share.`
`The backdrop trim was on for that run, and it dropped $72.2% of the sampled pixels — the blue cutting mat — before any color was counted. Turn the trim off and the mat comes back as the biggest swatch in the palette, which is exactly what you want when the “backdrop” is a color you plan to copy.`
`One more honest note on the numbers: $89.0% of the sampled pixels sit within a near match of one of the $6 swatches. The rest are the gradients between them — the shading down the armor, the fold in the cloak. That is expected, and it is why a palette is a starting point rather than a stencil.`
I would rather tell you the edges of a tool than have you find them halfway through a project.
If you want the whole beginner workflow — what to buy, in what order, and how to stop a first squad from looking like a test piece — that is what The Miniature Painting Level Up Guide is for. $17, 60+ pages, and no, you do not need it to use this page.
No, and it is worth being blunt about that. The page reads the colors of the pixels in your photo and nothing else: it has no paint database, no brand catalog and no idea what is on your rack. Use the hex codes and the hue families to narrow down where to look, then eyeball your own paints — or a paint rack chart — against the swatches.
Nowhere. The file is read by your own browser, drawn onto a canvas in the page and sampled in memory. It is never uploaded, never written to a URL, never stored in local storage and never sent to an analytics service. Close the tab and the pixels are gone.
Because the colors are cut into a fixed number of boxes. With four swatches the whole photo is described by four averages; with ten, the page keeps halving the widest-spread box, so narrow bands — a stripe, an eye lens, a gold trim — get their own box instead of being averaged away. More swatches is not more accurate, it is more detail.
It is the share of sampled pixels that ended up in that color's box. A swatch at 32% means roughly a third of the picture is that color. It is a coverage figure, so a backdrop or a large cloak will always outrank a gemstone, however interesting the gemstone is.
Yes — any photo works, and a painted color-scheme card is actually the cleanest input because the lighting is even and the swatches are flat. Artwork is harder: a painting has shadows, highlights and reflected color baked in, so the extract will describe the light in the picture as well as the local colors.
It averages the four corners of the photo and, if they agree with each other, treats that color as the surface the model is standing on and drops the pixels that match it before counting. It is why a miniature photographed on a blue cutting mat does not come back as a blue palette. If the corners disagree, nothing is trimmed.
They are ordinary sRGB hex values, so you can paste them into a design tool, a painting app or a notes file. What they are not is a print specification: this page does not do CMYK conversion, color management or ink matching, so do not treat them as press-ready values.