How it works
What gets smaller
Three passes run in order. First, objects that nothing in the file points to are dropped. Second, streams stored without compression are deflated, and the file is saved with compressed object streams. Third, if “Re-encode images” is on, each plain colour photo is decoded, scaled down to your longest-edge limit, and saved again as a JPEG at your quality.
A re-encoded image is only used if it is smaller than the original; otherwise the original stays. The result is only offered for download when the whole file is smaller than what you opened.
saved % = (before − after) ÷ before × 100
Strength
Light only repacks, so nothing visible changes. Balanced uses JPEG 75% and caps photos at 2400 px on the long edge, which suits screen reading and office printing. Strong uses 55% and 1600 px for email attachments; text inside images gets softer. Changing any control switches to a custom setting.
