Which sample size should I pick in the palette extractor: 100, 200 or 400 px?
Start at 200, and move to 400 when you can see a small area of color in the image that is missing from the palette. The tool shrinks the image so its longest side matches the sample size before counting colors, so 100 blends small details away while 400 keeps smaller areas such as a thin stripe or a small logo on the product.
What happens to your image at each size
A 4000 × 3000 phone photo is not analysed as it is. The tool first shrinks it with smoothing, then counts colors in the smaller copy:
- At 100 it becomes 100 × 75, just 7500 pixels. Each sample pixel stands for a 40 px wide square of the original.
- At 200 it becomes 200 × 150, 30000 pixels. Each sample pixel stands for a 20 px square.
- At 400 it becomes 400 × 300, 120000 pixels. Each sample pixel stands for a 10 px square.
An image smaller than the chosen size, such as a 150 × 100 icon with a 400 sample, stays 150 × 100 and is analysed as it is, because the tool never enlarges images.
Example: a white shirt with a thin red stripe
In that same 4000 × 3000 shirt photo, the red stripe is 8 px wide. After shrinking, it is 0.2 px wide in the sample at 100, 0.4 at 200 and 0.8 at 400. In every case the stripe blends with the white around it, but at 100 it all but vanishes into a faint pink, while at 400 it stays closer to its own color and carries more weight. That is why the red can be missing from the palette at 100 and show up at 400.
How to choose in practice
- Open the image in the Image palette extractor extractor and leave the sample size at 200 to begin with.
- Compare the palette with the image. Is there an important color in the picture that the palette lacks?
- If so, raise the sample to 400, and raise the number of colors if you need to.
- If the palette is full of near-identical shades of one color and you only want the overall impression, try 100: the stronger smoothing merges those shades.
- Read the analysis line under the palette: it states the actual sample dimensions and the number of pixels counted.
Is 400 always better?
Not necessarily. A bigger sample keeps more detail, including detail you do not want: camera noise, fabric grain, tiny reflections. Those add scattered colors that can take a slot in the palette. 100 suits the overall color mood of a busy image, such as a market scene or a landscape. 400 suits the case where one small touch of color matters to you.
Either way the result stays an approximation: each palette color is the average of a group of similar colors, and changing the sample size shifts those groups slightly.
Common questions
Is a larger sample slower?
A 400 sample has four times the pixels of 200, but the analysis runs in your browser and usually stays quick.
Does the sample size change my image file?
No. The shrunken copy is temporary and used only for the analysis; your image is untouched.
Is the sample size saved in the download?
Yes. The JSON file records the chosen sample size and the actual sample dimensions along with the other settings.