Drop an image here
or click to choose · or paste (Ctrl/Cmd + V)
processed in your browser · never uploaded
Turn any image's color scheme into English color words you can paste straight into an AI image prompt. Text-to-image models (Stable Diffusion, Midjourney, NovelAI and friends) largely ignore hex codes like `#3a7bd5`, but they respond well to color words such as `cobalt blue` or `teal`. This tool loads an image, extracts N dominant colors with median cut (a deterministic quantization, no randomness), and maps each one to the nearest entry in a built-in color-name table — about 70 words actually used in AI art (red, crimson, turquoise, beige, lavender and so on) — to assemble a ready-to-paste fragment like `crimson, navy blue, beige`. It's for when you want to borrow the palette of a reference image, a mood board or a photo for your own generation, and for when you're unsure how to put a color into words (is this pink salmon or hot pink?). A slider sets the number of colors from 2 to 12; when several dominant colors map to the same word you can merge the duplicates or keep them all. You can also weight the words by area, emitting AUTOMATIC1111-style `(navy blue:1.2)` so the most prominent colors carry more weight in the prompt. Each color is listed with a swatch, its English name, a Japanese name, the hex code and its area share, and you can click any swatch to copy just that one word. The color table and nearest-color math are pure computation bundled in the page — no external API or color database is contacted. The image you load is never uploaded, stored, or sent anywhere; it is processed only in your browser.
How to use
- Drop the image you want colors from (or click to choose, or paste with Ctrl/Cmd+V).
- Set how many colors to pull with the slider, and toggle "Merge duplicate words" or "Weight by area" if you like.
- Copy the resulting fragment like `crimson, teal, beige` and paste it into your generator's prompt box (or click a single swatch to copy just that word).
FAQ
Is the image I load sent to a server?
No. Both the color extraction and the mapping to color names run entirely in your browser, with no external API or color database. The image is never uploaded, stored, or sent anywhere — it is processed only on your device.
Why color names instead of hex codes?
Text-to-image models like Stable Diffusion and Midjourney barely understand hex codes such as `#3a7bd5`. Color words like `cobalt blue` or `teal` steer them far more reliably. This tool translates the dominant colors into English color words that actually work in a prompt.
Which names appear — are they exact colors?
Each output word is the closest match from a built-in table of about 70 color words (red, crimson, turquoise, khaki, lavender and so on) to the extracted dominant color. It's the nearest named color, not the exact color, so subtle in-between shades get snapped to the closest word. If you need the exact color, use the hex code shown on each swatch.
What does "Weight by area" do?
The more area a color covers in the image, the stronger a weight it gets, e.g. `(navy blue:1.2)` (the emphasis syntax used by AUTOMATIC1111, Forge and others). Use it when you want the dominant colors to push harder in the prompt. Turn it off to list plain color words.
How many colors should I pick?
Anywhere from 2 to 12. Use 3–5 to capture just the overall mood, or more to pick up small accent colors. Changing the count recomputes instantly without reloading the image.
Why do the same color words repeat — can I merge them?
When several dominant colors are similar they can map to the same word (e.g. two blues both becoming navy blue). Turn on "Merge duplicate words" to collapse duplicates and order them by prominence (largest area first). Turn it off to keep one entry per dominant color.