What the resize controls actually do
Resizing looks like one operation, but the controls on this page encode three separate decisions: how big, measured how, and what to do with the pixels in between.
Choosing the target
The three modes exist because “how big” has three different sources. A pixel target comes from outside — a form, a CMS, a print spec. A percentage comes from the file itself: you know it is roughly four times larger than you need. The longest-edge cap comes from the batch: twenty photos, some portrait and some landscape, all of which need to fit inside 2000 px without anyone deciding orientation by orientation.
The aspect ratio lock is on by default, and it should usually stay on. Typing a width and letting the height follow is the difference between a resized photograph and a stretched one. Hold Shift while dragging a handle when you genuinely want to distort — for a background texture, say, where nobody will notice.
What happens to the pixels
Every reduction throws pixels away. The question is only whether the discarded ones influence the survivors. A single-step downscale samples and discards; repeated halving averages, so edges stay continuous instead of breaking into stair-steps. That is why the tool reduces in stages, and why a 10x reduction here holds together better than the same reduction done in one jump elsewhere.
The resampling choice matters when the source is not a photograph. Smooth blending is right for anything with gradients and texture. Sharp sampling — nearest-neighbour — is right for pixel art, screenshots of low-resolution UI, and QR codes, where blending turns crisp squares into grey mush.
Working through a batch
Leave apply to all selected files on and the settings you tune on the visible image are used for every ticked file in the queue. The images are processed one at a time on your own device, so a batch of a hundred takes as long as a hundred single files — but you only make the decision once, and everything arrives as a single ZIP.
If the files vary a lot in size, longest-edge mode is the one that behaves sensibly across the whole set. Percentage mode will leave a 6000 px original at 3000 px while shrinking a 1200 px one to 600 px, which is rarely what you want when the destination is the same for all of them.