Scaling by a percentage is the right tool when you know roughly how much smaller you want something but not the exact pixel count — and it is much the better option for batches, because one percentage applies sensibly to images of different sizes while one pixel width does not.
Percentages apply to each side, not the area
This trips people up regularly. 50% means half the width and half the height, which is a quarter of the total pixels. 25% means an eighth of… no: a quarter of each side, so a sixteenth of the pixels. The area shrinks by the square of the percentage, which is why a modest-sounding reduction often cuts the file size far more than expected.
- 75% — a gentle reduction, roughly half the pixels. Useful when something is only slightly too big.
- 50% — the common choice. A quarter of the pixels, and typically a much smaller file.
- 25% — a sixteenth of the pixels. Good for turning camera photos into thumbnails.
- Above 100% — enlarging, which softens rather than sharpens. See below.
Where this beats exact dimensions
Batches, mostly. If you have thirty photographs at assorted sizes and orientations, setting them all to 800 pixels wide treats a panorama and a portrait identically, which is rarely what you want. Setting them all to 50% respects what each one already was. The same applies to a folder of screenshots from different displays.
It is also the natural fit for retina downscaling: an asset exported at 2× becomes its 1× version at exactly 50%, with no arithmetic on your part.
Going above 100%
You can, but the tool cannot add detail that was never captured. New pixels are interpolated from their neighbours, so a 200% enlargement is a softer, smoother version of the original rather than a sharper one. If you need something bigger, find the original file — a camera image, a vector, an untouched download — before resorting to an upscale.
Output format
The result is a PNG, because scaling is an editing step and PNG is lossless, so nothing is degraded by the resize itself. That does mean photographs come out considerably larger than the JPG you started with. If size matters, follow this with a compression step — in that order, so you are not scaling artifacts that were introduced first.
Everything runs in this tab on your own machine. Batches are processed one file at a time, which is slower than doing them at once but does not risk exhausting the tab’s memory partway through.