How to Turn a Photo Into Pixel Art

The three steps that make it work

Skip step one and you get a mosaic filter. Skip step three and you get a tiny image. Skip step two and you get blocky pixels with modern colour — a legitimate look, but not a retro one.

1. Downscale. Shrink the picture to the grid size you want — 64 pixels wide, say. The averaging is what turns detail into single meaningful pixels.
2. Quantise. Reduce how many colours are allowed. This is what makes the result read as retro rather than as a small photo.
3. Upscale with smoothing off. Blow it back up to a usable size using nearest-neighbour scaling so each pixel becomes a crisp square rather than a blur.

Choosing the grid size

At 64 pixels wide you have roughly 4,000 pixels in total to tell the whole story. That is about what a postage stamp shows at arm's length, and it is why the choice of photo matters more than any setting.

Pick a photo that can survive it

Crop before converting, not after. A tight crop is the single most effective thing you can do, because it changes how much of the tiny pixel budget lands on what you actually care about.

• One clear subject. Two competing subjects at 64 pixels become one blur.
• A simple background. Every pixel spent on clutter is a pixel not spent on the subject.
• Strong contrast between subject and background — in brightness, not just colour.
• Side or back lighting, so the shape reads when the detail has gone.
• No fine repeating patterns. Stripes and small textures alias badly at low resolution.

Why the colour limit is the retro part

Old hardware could not show many colours at once, and artists built a whole visual language inside that constraint. The original Game Boy managed four shades of green. The NES could display 25 colours on screen from a fixed palette of 54. Sprites were built from three or four carefully chosen shades per object, with dithering used to fake anything in between.

Dropping the colours-per-channel setting to four or six reproduces a lot of that feeling, because it forces the same kind of hard tonal steps. Leaving it high gives you blocky pixels in full colour, which looks modern rather than nostalgic.

Raise the contrast and saturation

Both survive downscaling better than fine texture does, and both help the subject separate from the background once detail is gone. Pushing saturation somewhat past natural is normal for this style — the quantisation step then snaps those pushed colours to distinct palette entries rather than muddy neighbours.

Always save as PNG

Pixel art is the worst possible input for JPEG. JPEG compresses by discarding high-frequency detail, and an image made entirely of hard edges between flat colours is nothing but high-frequency detail. You get grey ringing around every single square, immediately visible and impossible to remove.

PNG stores every pixel exactly, and because the image is large blocks of identical colour its compression works extremely well. A pixel-art PNG is frequently smaller than the JPG would have been.

Where it is useful beyond the aesthetic

• Placeholder art while a game is prototyped.
• Avatars and profile pictures that stay legible at tiny sizes.
• Anonymising a face in a photograph — though for genuine privacy, crop or black out rather than pixelate, because heavy pixelation can sometimes be reversed.
• Cross-stitch and bead patterns, where each pixel becomes one stitch or bead.
• Reducing a photo to a palette before turning it into a physical craft.