Why Compressing a PNG Often Makes It Bigger

Lossy and lossless are different machines

JPG, WebP and AVIF are lossy formats. They achieve their size by discarding information the eye is unlikely to miss, and a quality setting controls how much gets discarded. Ask for quality 60 and you get a smaller, slightly degraded image.

PNG is lossless. It stores enough information to reconstruct the original pixels exactly, and it compresses by finding patterns rather than by throwing anything away. There is no dial to turn, because there is nothing to trade. Pass a quality value to a PNG encoder and it is ignored — silently, which is the root of the confusion.

So why does it come back larger?

Because you have re-encoded the file with a different, usually weaker, encoder.

PNG compression works by prediction: for each pixel, the encoder guesses the value from neighbouring pixels and stores only the difference. How well this works depends on which prediction strategy the encoder tries and how much effort it spends searching. Dedicated tools test several strategies per row and pick the best. A browser's built-in encoder, which is what most web-based tools use, applies one general-purpose approach quickly.

If your original came out of a tool that optimised carefully, re-encoding it in a browser produces a file that is correct, identical in appearance, and bigger. Nothing has degraded. It is simply less cleverly packed.

What actually shrinks a PNG

Reduce the dimensions

This is by far the largest lever, and it is not really compression at all. Halving an image's width and height removes three quarters of the pixels. If a screenshot is displayed at 800 pixels wide but stored at 2400, resizing it with the image resizer will do more than any encoder setting could.

Reduce the number of colours

PNG supports a palette mode holding up to 256 colours. For logos, icons, diagrams and flat illustrations, this is often visually identical to the full-colour version at a fraction of the size. It is unsuitable for photographs, where the reduction is immediately visible as banding in smooth areas.

Change format, if you can

If the image is a photograph, PNG was the wrong container to begin with. Photographs have no repeating patterns for a lossless encoder to exploit, which is exactly the case PNG handles worst. Converting to WebP or JPG with the image converter typically produces a file several times smaller with no visible difference.

If the image needs transparency, WebP supports it and still compresses far better than PNG. There is rarely a reason to keep a transparent photograph as PNG in 2026.

When PNG is the right answer

PNG remains correct for a specific set of images, and it is worth keeping them in it:

• Screenshots containing text, where lossy compression produces visible artefacts around letterforms.
• Logos, icons and line art with sharp edges and flat colour.
• Images that will be edited repeatedly — each lossy save compounds the damage, while lossless saves do not.
• Anything requiring exact pixel fidelity: technical diagrams, medical or scientific imagery, colour references.

The general rule

Ask what the image is before asking how to compress it. A photograph belongs in a lossy format, and putting it in PNG is the actual problem — no compressor can fix a format mismatch. A screenshot or a logo belongs in PNG, and the way to shrink it is to reduce its dimensions or its palette, not to look for a quality slider that was never going to exist.

Frequently asked questions