Image Resizer

Making an image smaller discards pixels, and that information cannot be recovered — but a well-executed downscale looks sharp at its new size. This tool halves the image in stages before the final step precisely to avoid the aliasing that a single large resampling pass causes. Making an image *larger* is a different matter: there is no extra detail to invent, so enlargement always looks soft.

It keeps the proportions of the original. With the lock on, entering a width computes the matching height automatically, so nothing is stretched. Turn it off only when you genuinely need to force an image into a fixed box and are willing to accept the distortion — which usually looks wrong on anything containing faces or text.

For a full-width hero image, 1600–1920 pixels wide covers most displays without being wasteful. For an image inside a text column, 1200 is generous. For thumbnails, 400. A useful rule is to export at roughly twice the size the image will actually be displayed at, which keeps it sharp on high-density screens without doubling again.

Photographs: WebP if it is going on the web, JPEG if it needs to work everywhere. Screenshots, logos, diagrams and anything with sharp edges or flat colour: PNG, which keeps text crisp where JPEG would blur it. Anything with a transparent background must be PNG or WebP, since JPEG has no alpha channel and would fill the transparency with white.

Not here — resizing is a per-image decision, since the right dimensions depend on the source. If you want to apply one setting to a whole folder, the Image Compressor takes up to 20 files at once and can also convert their format.

No. Re-encoding through a canvas drops all metadata, including camera settings and any GPS coordinates. That is generally a benefit when publishing images publicly, but worth knowing if you rely on the metadata.

No. The file is read from disk directly into the page and processed by your browser in a Web Worker. Nothing is transmitted, and closing the tab discards everything.