Compress PNG images

What PNG is good at

PNG (Portable Network Graphics) was created in 1996 as a patent-free replacement for GIF. It is lossless: every pixel you decode is exactly the pixel that was saved. It supports full color, grayscale, indexed color with a palette, and a full alpha channel, so edges can fade smoothly into any background. That makes PNG the usual choice for screenshots, logos, icons, user-interface graphics, charts and anything with sharp text.

The price of being lossless is size. A photo saved as PNG is often several times larger than the same photo as a JPEG, because the noise and fine texture of a photograph cannot be predicted well enough to compress.

Why PNG files can still shrink

PNG compresses in two stages. First, each row of pixels gets a filter (None, Sub, Up, Average or Paeth) that replaces pixel values with their difference from neighbors; flat areas turn into long runs of zeros. Then the filtered data is packed with DEFLATE, the same algorithm as ZIP. Many programs pick filters and DEFLATE settings quickly rather than well, so the same pixels can often be stored in fewer bytes.

That gives two ways to make a PNG smaller:

• Lossless optimization. The tool runs OxiPNG, which tries different filter strategies and compression settings and keeps the smallest result. The image stays pixel-identical. Gains depend heavily on how the file was saved: an export from a quick screenshot tool often shrinks noticeably, a file that was already optimized hardly changes.
• Palette reduction (lossy). A full-color PNG with transparency stores 4 bytes per pixel. Converting it to a palette of up to 256 colors stores 1 byte per pixel before DEFLATE even starts, which is why this mode usually saves far more. Flat graphics and screenshots survive it well; smooth gradients and photos can show banding.

Palette reduction is the default ("Smart" mode, 256 colors with dithering); switch PNG compression to Lossless when every pixel must stay exact. The color count setting controls the palette size; fewer colors means a smaller file and more risk of visible steps in gradients.

Settings for common jobs

Retina screenshots are a frequent source of huge PNGs: a full-screen capture on a 2880 × 1800 laptop display has more than 5 million pixels. Resizing the longest side to 1600 px cuts that to less than a third and usually matters more than any compression setting.

For orientation: in our tests, four PNG screenshots of Wikipedia, MDN and GitHub pages became 58–60 % smaller with the default Smart mode and, at the median, about a third smaller in Lossless mode. Your own files may shrink more or less.

Common problems

• Banding in gradients after Smart mode. Raise the color count or switch back to lossless.
• A jagged halo around a transparent logo. Antialiased edges need several levels of transparency; keep at least 128 colors for such images.
• Almost no saving in lossless mode. The file was already well packed. Only palette reduction or smaller dimensions will make a real difference, and the tool returns the original file when nothing smaller can be made.
• The PNG is a photograph. JPEG or WebP fits the content better and will be far smaller.

PNG questions

More image tools

The image compressor handles a mixed batch of files. Animated graphics belong in the GIF compressor, and how it works explains what lossless and lossy mean in more detail.