Drop a picture here or click to choose
Or load the sample. It is drawn and dithered on a canvas in this tab: your files and inputs never leave your device.
No image loaded yet. Pictures wider or taller than 2048 pixels are scaled down first, and the preview is magnified with hard pixel edges so every dot the algorithm placed stays visible.
Related tools
- Image SteganographyHide a text message or a whole file inside an image's pixel bits, then extract it back out, entirely in your browser.
- Image to ASCII ArtConvert any image into ASCII text art, colored ANSI art, or Unicode braille dot art.
- Redaction ToolBlack out parts of a screenshot by destroying the pixels, not hiding them.
- Meme GeneratorAdd classic caption text to any image and export it as a PNG, with no account and no watermark.
- Photography CalculatorsDepth of field, hyperfocal distance, exposure value, ND filter times, and field of view in one text box.
- EXIF Viewer and StripperSee every EXIF, XMP and IPTC field in a photo, then export a clean copy with all of it removed.
What it does
Converts any image to a fixed palette using the dithering algorithm you pick, from the seven classic error diffusion kernels (Floyd Steinberg, Atkinson, Jarvis Judice Ninke, Stucki, Burkes, Sierra, Sierra Lite) to ordered Bayer matrices at 2x2, 4x4, and 8x8, plus a 64x64 blue noise tile, plain thresholding, and seeded random noise. Palettes cover 1 bit black and white, 4 and 16 level grayscale, the Game Boy greens, CGA, Commodore 64, PICO-8, the black white red e-ink combination, the 7 color ACeP e-paper set, and any list of hex colors you paste in. All the math runs in linear light by default, so a dithered midtone reflects the same amount of light as the original instead of coming out too bright. A pixel scale option shrinks the image with a box filter before dithering, which is how you get clean chunky pixel art rather than a fine dither pattern squeezed into a small canvas.
How to use it
Drop or pick an image, then choose an algorithm and a palette. The preview redraws as you change anything, and the result is upscaled with hard pixel edges so you can actually see what each algorithm is doing. Turn the pixel scale up to shrink the image before it is dithered, drag the strength down if the pattern is too loud, and switch serpentine scanning off if you want the older left to right look with its diagonal streaks. Download the PNG at its dithered size when you are happy with it.
Why this one
Most dithering sites either upload your picture to a server or give you one algorithm and one palette, which is useless when the whole point is comparing Atkinson against Floyd Steinberg on your own photo. This one runs 13 algorithms and 10 palettes in the tab you already have open, so your files and inputs never leave your device, and it does the arithmetic in linear light, which most online converters skip and which is the reason their output looks washed out. There is no account, no watermark, no size cap other than your own memory, and the page keeps working with the network off.
FAQ
- Which dithering algorithm should I use for an e-ink or e-paper display?
- Start with Floyd Steinberg on the black white red or 7 color ACeP palette. Error diffusion suits e-paper because the panel holds a still image and nothing crawls between frames, and Floyd Steinberg gives the most detail per palette color. If the picture comes out muddy on a 7 color panel, try Atkinson: it drops a quarter of the error on purpose, which lifts contrast and stops flat areas from turning into noise. Avoid Bayer for photographs on e-paper, since the visible grid competes with the panel's own dot structure. Blue noise is the one ordered pattern worth trying, because it has no repeating structure to clash with.
- Why does my dithered image look darker than the original?
- Because it is correct. sRGB is not linear: a pixel of 128 emits about 21 percent of the light of white, not 50 percent. If you dither in sRGB numbers, half your pixels turn white and the result reflects 50 percent of the light, so it looks too bright. Doing the math in linear light lights up about a fifth of the pixels instead, which matches what the original actually emitted. Most other converters skip this step, so their output looks brighter and yours looks right. If you want the old look on purpose, turn off gamma correct.
- Can I export the result for a Game Boy screen or an e-paper panel?
- Yes. Set the pixel scale so the output lands at the native resolution of the target (160 by 144 for a Game Boy, whatever your e-paper panel reports), pick the matching palette, and download the PNG. The download is the dithered image at its real pixel size, not the magnified preview, so every pixel in the file is one pixel on the device and every color in it is one of the palette entries. From there you can feed it to your usual tile converter or panel driver without a second quantization step undoing the dither.
Keyboard shortcuts: press ? anywhere on this page to see them.