How to use the Image Posterize Tool
- Open an image. Drop a photo, painting or sketch onto the box, paste it with Ctrl+V, or click to browse. The sunset sample shows what a five-color split looks like before you load your own.
- Pick a method. Smart colors finds the most representative colors in the picture. Channel levels cuts red, green and blue into equal steps. My inks maps the image onto colors you already own — type HEX codes or tap a preset such as Red + black.
- Clean it up. Raise Smoothing (median filter) to flatten film grain and JPEG noise, and Remove specks smaller than to merge tiny islands into their neighbors. Drag across the preview, or use the Compare with original slider, to check what was lost.
- Tune the palette. Click any swatch to set an exact color — it gets locked and the remaining colors re-fit around it. The lock button keeps a color fixed while you change the count or hit Try another split.
- Export. Download PNG saves the full-size posterized image. Separations (ZIP) gives one black-and-white layer per screen, the composite, and a
palette.txtwith print order and coverage. Each separation also has its own PNG button.
Why fewer colors print better
A phone photo can use any of 16,777,216 possible RGB values, and sensor noise alone means a single shot usually contains tens of thousands of distinct ones. A screen print, linocut or stencil can’t work that way. Every color is a separate screen to coat, expose, register and wash out, so each one adds setup time and a new chance for misregistration. That is why hand-pulled posters are planned around a short list of inks.
Posterizing makes that decision visible before you burn a single screen. It also helps painters: squinting at a reference reduced to five flat shapes shows you the big value masses far more clearly than the original. If value is what you’re after, the Notan & Value Study Tool goes further in grayscale.
The two main methods behave very differently. Channel levels is the classic Photoshop Posterize command: each RGB channel snaps to n evenly spaced values, so the total palette is n³ — 8 colors at 2 levels, 27 at 3, 64 at 4, 125 at 5 and 216 at 6. It is fast and predictable, but it spends colors on combinations your image barely uses and can turn a soft sky into stripes of unrelated hues.
Smart colors runs k-means clustering in CIELAB, a color space built so that equal distances look roughly equally different to the eye. It samples up to 24,000 pixels, starts with well-spread seeds, and moves each cluster to the average of its members until they stop shifting. Five clusters chosen this way usually beat 27 channel levels for a print, because every ink earns its place.
How the speck filter translates to paper
Remove specks smaller than is measured on the preview, which is up to 1,100 px on its long side, and scaled up automatically for the full-size export. For any image at least that large, here is roughly what each setting removes once printed:
| Setting | About the size of | On a 24 in (61 cm) long poster | On an 11 in (28 cm) long print |
|---|---|---|---|
| 4 px | 2 × 2 px square | 1.1 mm | 0.5 mm |
| 16 px | 4 × 4 px square | 2.2 mm | 1.0 mm |
| 36 px | 6 × 6 px square | 3.3 mm | 1.5 mm |
| 81 px | 9 × 9 px square | 5.0 mm | 2.3 mm |
| 196 px | 14 × 14 px square | 7.8 mm | 3.6 mm |
| 400 px | 20 × 20 px square | 11.1 mm | 5.1 mm |
The filter works on connected regions, so a 16-pixel sliver one pixel wide counts the same as a neat 4 × 4 dot. A lone speck gets the color that surrounds most of its edge, which keeps outlines intact rather than smearing them. For hand-cut stencils and vinyl, aim higher: anything under about 3 mm is fiddly to weed.
Getting separations ready for the screen
The separations come out in print order, lightest to darkest — the usual sequence because a dark ink covers small registration errors in a lighter one underneath. With Lightest color is the paper ticked, that color gets no screen and is labeled Paper instead; choose a stock close to it. If the image already has a transparent background, the transparency plays the paper and every color gets a screen.
Black on white (film positive) is what you print onto transparency film for exposing emulsion. Black on transparent drops straight into a layer stack, and Ink color on transparent lets you rebuild the print digitally to check overlaps. Add registration marks puts identical crosshairs outside the artwork on every layer, so the films line up on the light table. Everything is computed on your device; the image never leaves the browser.
Need a tone that a flat color can’t give? Run that separation through the Halftone & Dot Effect Generator to turn it into printable dots.
Tips from practice
Lock the paper color first. Click the lightest swatch, set it to your actual stock (cream cover stock is rarely pure white), and the remaining inks re-fit around it. The result previews far closer to the real print.
Smooth before you reduce, then despeckle. A median filter of 1–2 px removes sensor noise without rounding off corners; the speck filter then handles whatever islands survive. Pushing smoothing to 5 px makes lettering go blobby.
Use lightness order for odd ink sets. With My inks and Lightness order, the darkest cluster always gets your darkest ink, whatever its hue. That is how you get a convincing teal-and-orange print from a gray winter photo.
Count screens, not swatches. Two near-identical browns are two screens and two print passes. If the swatches look alike, drop the color count by one and see whether anything important disappears.
Check thin lines at full size. The preview is downscaled. Open one separation PNG at 100% before you send it to film, especially hair, text and fine outlines.