Compress

Compress PNG Images

Two ways to shrink a PNG: lossless optimisation that keeps every pixel exactly as it was, or colour reduction that typically cuts file size by 60-80%. Transparency survives both.

Compression

Compression type
3
Higher searches harder for a smaller encoding. Above 4 costs a lot of time for very little gain.
Lossless: every pixel comes out exactly as it went in. Expect a saving of roughly 10-30%. That is the honest ceiling for lossless PNG — anything advertising much more is reducing colours.
Your files never leave this device. Everything happens inside your browser, so there is no upload, no server copy and nothing to delete afterwards. You can test that claim: load this page, turn off your internet connection, and the tool still works.

Two different things called "PNG compression"

PNG is a lossless format. Its compression works the way a ZIP file does: it finds repeated patterns and stores them more efficiently, and the pixels you get back are exactly the pixels that went in. That places a hard ceiling on how much can be saved without changing the image, and that ceiling is usually somewhere between 10% and 30%.

So when a website advertises 70% savings on PNG files, it is not doing lossless compression. It is reducing the number of colours in the image — replacing thousands of subtly different shades with a palette of a few hundred — and calling the result compression. That technique is genuinely excellent for the kind of images PNG is designed for, but it does change the picture, and you deserve to know which one you are getting.

This tool separates them:

  • Lossless runs the file through OxiPNG, which searches for a more efficient encoding of identical pixel data. Typical saving: 10-30%. The output is pixel-for-pixel identical to the input.
  • Reduce colours rebuilds the image from a smaller palette using median cut, with optional dithering. Typical saving: 60-80%. The image changes, usually imperceptibly on graphics.

When colour reduction works, and when it does not

Colour reduction is close to free quality-wise on images that only ever contained a handful of colours in the first place. A logo with four brand colours, a screenshot of a user interface, a flat illustration, a chart, an icon — all of these can drop to 64 or even 32 colours with no visible difference at all, while the file collapses to a fraction of its original size.

It works badly on photographs. A photograph contains tens of thousands of distinct colours in every smooth gradient, and forcing those into 256 slots produces visible banding across skies and skin. If your PNG is a photograph, the honest answer is that it should not be a PNG at all — JPG or WebP will be several times smaller at better quality.

What dithering does

When you reduce a smooth gradient to a small palette, the gradient turns into visible stripes, because a range of shades that used to be continuous now has to snap to a few fixed values. Dithering scatters pixels between the two nearest available colours in a fine pattern, and from normal viewing distance your eye blends them back into the shade that is missing.

The cost is file size: that fine scattered pattern is noisier and therefore less compressible than a flat area, so a dithered PNG is typically 10-20% larger than an undithered one at the same palette size. Turn it on for anything with gradients, soft shadows or subtle lighting. Turn it off for flat graphics, where it only adds noise to areas that were perfectly smooth to begin with.

Transparency is preserved

Both modes keep the alpha channel intact, including partial transparency. That matters for anti-aliased edges: the smooth boundary of a rounded logo is made of semi-transparent pixels, and a tool that flattens alpha to fully-on or fully-off leaves you with visibly jagged edges.

If you need to remove transparency deliberately — because a marketplace requires a solid white background, for instance — use the PNG to JPG converter, which lets you choose the colour that replaces it.

How to compress PNG Images

  1. Drop your PNG files onto the box above.
  2. Pick Lossless to keep every pixel identical, or Reduce colours for a much smaller file.
  3. If you chose colour reduction, adjust the colour count and watch the preview until the quality is acceptable.
  4. Download the results individually or as a ZIP.

Limits worth knowing

Every tool has boundaries. These are this one's, stated plainly so you do not waste time discovering them yourself.

  • Lossless mode typically saves 10-30%. If you need dramatic savings you have to use colour reduction, which does change the image.
  • Colour reduction works beautifully on logos, screenshots, icons and flat illustrations. It works poorly on photographs, which are better served as JPG or WebP.
  • A photograph saved as PNG will almost always be far larger than the same photograph as JPG, no matter how hard you optimise it.

Frequently asked questions

What is the difference between the two modes?
Lossless mode reorganises how the file stores its data, finding a more efficient encoding of exactly the same pixels. The output is pixel-for-pixel identical to the input. Colour reduction rebuilds the image using a smaller palette, typically 256 colours or fewer, which is a much bigger saving but does alter the image.
Why is lossless PNG compression only saving 15%?
That is normal and it is the honest ceiling. Tools that advertise 70% savings on PNG are almost always doing colour reduction and calling it compression. This tool separates the two so you know which one you are getting.
Does transparency survive compression?
Yes, in both modes. Lossless mode preserves the alpha channel exactly. Colour reduction preserves alpha too, including partial transparency such as soft shadows and anti-aliased edges.
Should my photo be a PNG at all?
Usually not. PNG is designed for graphics with flat areas of colour and hard edges: logos, screenshots, diagrams, illustrations, anything needing transparency. For photographs, JPG or WebP will be several times smaller at the same visible quality.
What is dithering and should I turn it on?
When colours are reduced, smooth gradients turn into visible bands. Dithering scatters pixels between two nearby colours to fake the missing shades, which hides the banding at the cost of a slightly larger file. Turn it on for anything with gradients or soft shadows, and off for flat graphics where it just adds noise.