Fourteen dither algorithms, from Floyd-Steinberg and Atkinson to Bayer and blue noise, plus retro palettes. Compare them on your own photo. Nothing is uploaded.
Open the editor with this effect
There is no generative model behind this look and nothing was trained on anyone's work. It is a short list of fixed steps running in your browser.
Dithering lets an image with only a few colours pretend it has hundreds. Instead of rounding every pixel to the nearest colour and getting flat bands, it scatters dots of the colours you do have so your eye blends them back into tone. This page is the map: every dither method Bitgrain has, what each one looks like, and when to reach for it. Drop in a photo, try them side by side, and nothing ever leaves your device.
Open the editor from this page and drop in a photo. It loads with Floyd-Steinberg already on, in black and white, so you see a result straight away.
Set the resolution slider first. It decides how big one dither dot is compared with the picture. Lower gives chunky, visible grain, higher gives a fine print texture. Most muddy results come from skipping this step.
Lift the contrast next. Dithering has very little room for tone, so a slightly punchier photo going in almost always gives a cleaner result coming out.
Switch methods in the dither panel until one suits the subject, add a palette if you want colour, then export a PNG. The image is processed in your browser, so it is never uploaded anywhere.
Error diffusion rounds each pixel to the nearest colour, then passes the rounding error on to pixels it has not reached yet. The methods differ in how far that error travels and how much of it is kept. Wider spread means smoother tone and less visible pattern. Narrower spread means crisper, busier grain.
Floyd-Steinberg passes all the error to four neighbours. Even, fine grain that keeps the most detail. The safe default for photos and anything you want to stay recognisable.
Atkinson passes on only three quarters of the error and throws the rest away. Highlights clip to clean white and shadows fill in, so the picture gets brighter and more graphic. Use it for portraits in good light and for small screens.
Stucki spreads the error across twelve neighbours over three rows. Very smooth tone with sharp edges, good for portraits and for large black and white prints.
Jarvis-Judice-Ninke also reaches three rows but weights the far pixels more. The softest gradients and the least visible pattern of the set, at the cost of a little crispness.
Burkes covers two rows with seven neighbours. It looks close to Stucki and is a good middle ground when Stucki feels too soft.
Sierra covers three rows, Sierra two-row covers two, and Sierra lite touches only three neighbours. Run them in that order and the grain goes from soft and clean to tight and slightly busy.
These methods compare every pixel with a fixed pattern of thresholds instead of passing error around. Because the pattern belongs to the position and not to the picture, it stays put when the image changes, which matters for animation and for anything printed in a series.
Bayer 2x2 is coarse and unmistakably digital, with only a handful of tones. Use it when you want the pattern to be the point.
Bayer 4x4 is the classic console look, a regular crosshatch that reads clearly at small sizes. Pair it with a four colour palette for handheld game screens.
Bayer 8x8 is a finer grid with more tones. It reads as texture rather than as a pattern from a normal distance and holds up well on posters.
Blue noise gives an even, organic speckle with no visible grid. It sits between error diffusion and Bayer: no repeating pattern, but just as stable from frame to frame.
Threshold is pure black or white with no texture at all. Use it for logos, text, line art and stencils.
Random scatters dots with no structure, which looks like static or a photocopy of a photocopy. Use it for grit, not for detail.
Beyond the dither panel there is a Dither look in the look picker, built for retro and print styles. It adds patterns the classic panel does not have: a Bayer 16x16 grid, a clustered dot screen like a newspaper halftone, a radial pattern you can rotate, horizontal, vertical and diagonal line screens, and white or blue noise.
It comes with its own palettes: Mono 1-bit, grey in four or eight levels, Game Boy, both CGA palettes, Pico-8, C64, NES, 3-bit RGB, Cyberpunk, Pastel, Risograph, Sepia, or the photo's original colours. Pixel size sets how chunky the result is, and colour matching lets you dither by brightness only or one channel at a time for a printed, slightly separated look.
For e-ink screens, receipt printers and laser engravers, start with Atkinson or Stucki at the device's exact pixel size. Fine diffusion grain tends to blur on hardware that bleeds dots together, so drop the resolution a step if the print looks grey.
For posters and prints, Bayer 8x8 or Floyd-Steinberg at a low resolution gives dots big enough to survive the trip to paper. Export at 2x or 4x on Pro and the dots stay square instead of going soft.
For video, loops and anything animated, use an ordered method or blue noise. Error diffusion recalculates the whole pattern every frame, so the grain crawls. Ordered patterns stay in place.
For faces and skin, Stucki or Jarvis-Judice-Ninke keep gradients smooth. For a bold, graphic portrait, Atkinson. For a retro game look, Bayer 4x4 with one of the twenty four hardware and print palettes.
Yes, you can try every dither method without an account. Free exports carry a watermark and stop at 1080 pixels. Pro removes it and exports at full resolution, up to 4x.
No. Every method runs in your browser on your own device. The photo is never sent to a server, and the same settings always give the same pixels.
There is no single best. Floyd-Steinberg keeps the most detail, Atkinson looks the most graphic, Stucki and Jarvis-Judice-Ninke are the smoothest, and Bayer gives a steady pattern that suits print and animation. Try three on the same photo and keep the one that suits it.
Yes. Turn colour on and pick a palette, like Game Boy, CGA, C64, NES, Pico-8, Risograph or Sepia, and the same methods dither the photo into those colours instead of black and white.
Usually the resolution is too high, so the dots are too small to see. Lower the resolution so each dot covers more of the picture, then raise the contrast a little before you export.
Dither exports as PNG. The dots have hard edges, so a 2x or 4x export on Pro enlarges cleanly for large prints. Halftone and ASCII looks are the ones that export as true vectors.