Why an Image Sometimes Gets Larger After Compression

Re-encoding an image doesn't always remove bytes. Three common reasons it grows:

A worked example

Take a 2.8 MB JPG straight from a phone. Camera apps already save at a fairly high quality, so dropping it into Image Compressor on the Best qualitypreset (92%) can produce a 3.1 MB file: higher than the camera's own setting, with nothing gained. The compressor notices, keeps your original and labels the row "kept original (compression didn't help)". Switch to Balanced(75%) and the same photo commonly lands well under half its starting size, because 75% is below the point where the camera encoded it and the encoder finally has something to discard.

The PNG case is more dramatic. A 4000 × 3000 photo exported as PNG by a screenshot or editing app can weigh 14 MB, and re-encoding it as PNG leaves it at 14 MB or nudges it higher. The compressor keeps the format on purpose, so the fix is a conversion: the same picture as JPG at 80% is typically under 2 MB, and WebP smaller again.

The quality slider is not a size slider

A quality of 75% does not mean 75% of the bytes. The relationship is steeply non-linear: the step from 100% to 90% can halve a file, while the step from 75% to 60% might only shave off a fifth, and visible artefacts grow faster at the low end than savings do. If you need a guaranteed number — a form that insists on 500 KB, say — the compress-to-target tools search the quality range automatically until the output fits, instead of guessing.

When pixels, not quality, are the problem

File size scales with the number of pixels far more reliably than with the quality setting. Halving both the width and the height of a 4032 × 3024 photo leaves 2016 × 1512 — one quarter of the pixels, and roughly a quarter of the bytes at the same quality. For anything viewed on a screen, 1920 pixels on the long edge is more than most displays can show, which is why Image Studio's fit within resize mode is usually a bigger win than any slider. It never enlarges a smaller image, so there is no risk of the opposite problem.

Two deliberate exceptions. If you untick that option, the compressor still hands you the larger file but flags it as larger than the original rather than reporting a saving. And Remove Image Metadata never keeps the original at all, because the original is the file that still contains the metadata; it re-encodes at 95% quality and accepts a slightly bigger output as the cost of a clean one.