Image Optimizer

The quality slider controls how aggressively the encoder discards visual information. It is not a percentage of the original file size, and the relationship between the two is not linear.

In practice the curve has a knee. Between roughly 100 and 80 the file shrinks substantially while the difference is hard to see on most photographs. Below about 60 the losses become visible: soft blocking in flat areas like skies, and coloured fringing around hard edges and text. Where exactly that happens depends on the picture — a photograph tolerates far more compression than a screenshot of an interface.

Re-compressing an already-compressed image compounds the damage, because the encoder is discarding information from a picture that has already had information discarded. Always optimise from the best original you have rather than from a file that has been through this once already.

One consequence is worth knowing before you use the result. Processing happens by drawing the image onto a canvas and encoding what was drawn, and a canvas holds pixels only. Every piece of metadata is therefore dropped: EXIF, the camera model, the timestamp, the colour profile, and any embedded GPS coordinates. That is excellent if you are stripping location data before posting a photo, and a real loss if you were relying on capture dates to keep an archive in order. Keep the original.