Image Batch Resizer

Resizing one image is easy; resizing fifty is a chore, and doing it by forcing every file into the same rectangle wrecks half of them. This Image Batch Resizer takes one sizing rule and solves it separately against each image's own dimensions, so a folder mixing portrait and landscape shots comes out consistently sized rather than squashed.

Nine modes cover the ways a batch actually needs to be sized: fit inside a box, fill and crop to it, pad to it, stretch to it, or work from a single number — longest edge, shortest edge, an exact width, an exact height, or a percentage of each image's own size. Photographers preparing a gallery, sellers standardising product shots and developers trimming an asset folder all run the whole set in one pass, watch the projected output beside each queued file, and take the results as a ZIP.

The distinction that matters in bulk work is between a target size and a sizing rule. A target size suits a batch where every source already shares an aspect ratio; anything else needs a rule that adapts. Fit within scales each file until it sits inside your box, so outputs vary slightly in size but nothing is cropped or distorted. Fill and crop guarantees identical dimensions by scaling to cover the box and trimming the overflow toward the anchor you pick. Pad to exact also guarantees identical dimensions but fills the leftover space with a colour instead of cutting anything away.

Longest edge is usually the right answer for a camera roll: cap the long side at 1600 pixels and portrait and landscape frames both end up with the same amount of detail, each keeping its own shape. Percentage works from each file's own dimensions, so a 4000-pixel photo and an 800-pixel one shrink by the same proportion rather than converging on one size.

Reductions are done in stages. Halving repeatedly until the last step is a small one avoids the sparse sampling that makes a single large downscale look gritty on detailed photographs, which is the difference you notice when a 4000-pixel image lands at 400.

Output can keep each file's own format or convert the batch to JPEG, PNG or WebP. Browsers cannot write GIF at all, and most cannot write BMP or AVIF, so those are saved as PNG and labelled in the results rather than handed back with an extension that lies about the bytes. Transparency needs a moment's thought too: JPEG has no alpha channel, so a transparent source is flattened onto the background colour, and white is used if you asked for transparency anyway.

The size budget is the other half of a web-prep run. Give it a ceiling in kilobytes and each file is encoded at the highest quality that fits underneath, searched per image rather than guessed once for the batch. Anything that cannot get under the ceiling even at minimum quality is flagged, which is your cue that it needs smaller dimensions rather than harder compression.

Two things do not survive. Re-encoding discards metadata, so GPS coordinates and camera settings go — welcome for holiday photos, less so if captions or copyright fields mattered. And animation is flattened to a single frame, since the drawing surface only ever holds one. The exception is a file the settings leave completely untouched: those are passed through byte for byte, animation and metadata intact, and marked as unchanged in the results.