Palettes from pictures

Suisai can extract colors from a photograph or image and turn them into an editable palette.

Click the picture button in the palette’s control row, or drop an image file onto the palette window. Suisai displays the image above the swatches and extracts colors using one of three methods: Dominant, Vibrant, or Spread.

You can adjust the number of extracted colors, choose individual colors with pins, or open the picture in a larger window for closer inspection.

The picture preview can occupy one-tenth, one-quarter, or one-half of the palette window’s height. Choose the size from the menu beside the filename.

The picture is never kept

Suisai processes photographs entirely on your Mac. An image is loaded into memory while you’re working with it, but Suisai doesn’t save a copy, thumbnail, or path to the original file.

Saving a palette saves its colors, not the picture.

When you dismiss the image or quit Suisai, the image is no longer retained. Even if you reopen the palette using Where I Left Off, the colors return without the original picture.

Nothing is uploaded, and no account is required.

Three ways to extract colors

Suisai can automatically extract between 1 and 16 colors from an image. Each extraction method approaches the picture differently.

• Dominant identifies the colors covering the largest areas. For a landscape with broad areas of sky, water, or ground, it reflects the main colors making up the scene.
• Vibrant prioritizes colorful areas that occupy a meaningful portion of the picture. Isolated bright pixels are less likely to be selected than saturated colors that occur throughout a region.
• Spread looks for a wider range of distinct colors. Rather than returning several similar blues from a harbor photograph, it can identify contrasting colors elsewhere in the image.

The methods work from groups of related pixels, rather than simply averaging the entire picture. Selected colors are matched to actual pixels in the image, so each result corresponds to a color that exists somewhere in the original.

Use the + and − controls to change the number of automatically extracted colors, up to a maximum of 16.

Changing the method or count updates the palette while preserving the manual selections you’ve made.

Pins

Every automatically extracted color has a pin showing where its color came from in the image.

Click anywhere on the picture to add a manual pin and sample the color at that position. Option-click a pin to remove it.

Manual pins are distinguished by a second ring. They remain in place when you change the extraction method or the requested number of colors.

You can also move an existing pin or edit an extracted color. Doing either makes that selection manual, so Suisai preserves it during subsequent extractions.

The automatic extraction limit is 16 colors, but an image supports up to 32 pins in total, including manually placed selections.

This lets you keep particular colors you’ve chosen while Suisai recalculates the remaining selections around them. Manual pins take priority over automatically generated ones.

The picture window

Choose Enlarge to open the picture in a larger window. This makes it easier to inspect fine details and place pins precisely.

Use these controls to navigate:

The enlarged window also contains the Dominant, Vibrant, and Spread controls, together with the extraction count. You can make selections and refine the palette without returning to the smaller preview.

Choose Replace to open a different image, or Done to return to the palette.

How extraction works

You don’t need to understand the extraction process to use it. The following explains why the three methods produce different results and why repeated extractions are consistent.

Preparing the image

Suisai first reduces the image to a maximum of 160 pixels along its longest side.

This provides enough information to identify the main color regions without processing every pixel in a high-resolution photograph. It also helps reduce the influence of sensor noise and JPEG artifacts.

Pixels that are mostly transparent are ignored.

Grouping similar colors

Suisai groups colors according to their perceptual similarity rather than their numerical RGB values.

Two RGB values can be numerically different yet appear almost identical. Conversely, relatively small numerical differences can produce colors that look noticeably different.

Working in a perceptual color space helps Suisai form groups that better reflect the colors a person sees.

The extraction process initially produces approximately three times as many groups as the requested palette size. This gives Suisai a larger selection of candidates from which to choose the final colors.

Selecting real colors from the picture

Once the groups have been identified, Suisai selects a real pixel near the center of each chosen group.

This is why extracted colors have corresponding pins: they identify locations in the original image.

The selected colors aren’t merely mathematical averages that may never have appeared in the photograph.

Consistent results

The grouping process needs starting points. If those points are selected randomly, repeating the extraction can produce slightly different palettes.

Instead, Suisai chooses its starting points from a histogram of the image’s colors, favoring well-populated regions while keeping the starting colors separated.

This makes extraction deterministic: the same image and settings produce the same result, rather than changing randomly between runs.

How each method chooses its colors

After grouping, each extraction mode selects its candidates differently.

Dominant chooses the groups representing the largest portions of the image.

Vibrant ranks candidates by colorfulness. It also requires a group to represent at least 1% of the image and excludes colors below its chroma threshold. This prevents isolated bright pixels or relatively neutral areas from dominating the result.

Spread begins with the largest group, then repeatedly chooses the candidate most different from those already selected. This encourages a broader range of colors across the palette.

Dominant and Vibrant also suppress near-duplicates, giving greater importance to meaningful differences in hue and saturation rather than treating every change in lightness as a separate color.

Why automatic extraction stops at 16

Automatic extraction is intended to identify useful, distinct colors from an image.

As the requested count increases, Suisai must divide the image into progressively finer groups. Eventually, additional colors may differ so little that they contribute relatively little to the palette.

The 16-color limit balances useful detail against unnecessary repetition.

Manual pins serve a different purpose. They represent colors you’ve deliberately selected, rather than groups discovered by the extraction algorithm. That’s why you can continue adding them beyond the automatic extraction limit, up to 32 pins in total.