Compress a JPEG

A JPEG is stored as a grid of blocks, each 8 by 8 pixels. For every block the encoder keeps the broad tones exactly and rounds off the fine detail, and the quality number sets how coarse that rounding is. Colour is treated more roughly than brightness, because eyes are far less sensitive to it. That is why a portrait survives heavy compression while red lettering on a blue background falls apart quickly.

We encode with MozJPEG settings: progressive scans and tables tuned to each picture. At the same quality number the JPG is usually a few percent smaller than the one a camera or a phone writes, and you will not see the difference.

Smart is the default on the home page and the safe choice when you do not want to think in numbers. It compares candidates against the original and keeps the lowest setting, between 60 and 88, that still looks the same. Quality gives you the slider, and the same number for every file in a batch. Target size works the other way round: you name a limit, say 200 KB, and the tool finds the best quality that fits under it.

There is a Lossless mode too, but JPEG has no lossless encoding. Choose it for a JPEG and you get your own file back untouched, with a note that says so.

Every save as JPEG rounds the detail off again. A photo that went through a messaging app, then an editor, then a compressor has lost a little at each step, and the damage adds up as smeared textures and halos around edges. Keep the file that came off the camera and compress that one. If an already small JPG is all you have, only a quality below the one it was saved at will shrink it. A higher number makes the file bigger and brings nothing back.