Resize image

Choose images, type a width and height or drag a percentage, and get them back at that pixel size. The line under the controls shows the exact output dimensions before you run anything, which is the part most resizers leave you to guess at.

Resizing and compressing are not the same job

Compressing changes how many bytes are used to describe the same grid of pixels. Resizing changes how many pixels there are. Both make a file smaller and they are routinely confused, but only one of them is the right answer to a given problem.

If a form says "maximum 2 MB", either will work and compressing usually costs less quality. If a form says "maximum 1000 × 1000 pixels", or a website is loading slowly because a 4000-pixel photo is being displayed in a 400-pixel box, compression cannot help you at all — you need fewer pixels.

The second case is worth dwelling on, because it is the most common invisible waste on the web. A browser shown an oversized image must download all of it, decode all of it into memory, and then throw most of it away while scaling down. Resizing to the size it is actually displayed at is often a bigger win than any amount of compression.

Aspect ratio: what "locked" actually does here

With the ratio locked and only a width given, the height follows automatically. Give both and the image is fitted inside that box — the result may be narrower or shorter than what you typed, but it is never distorted and never cropped. This is the behaviour you want almost always.

Unlock the ratio and the image is stretched to exactly the numbers you give. There are legitimate reasons to want that — undoing a known distortion, matching a fixed banner slot where a little stretch is acceptable — but people who reach for it by accident get subtly wrong-looking faces.

What this tool deliberately does not do is crop to fill a box. Cropping is a decision about what to throw away, and it needs a preview with a draggable frame to be made properly. Silently cutting off the sides to make a photo fit a square would be the wrong kind of helpful.

Enlarging, and the honest limits of it

You can enter dimensions larger than the original, and the tool warns you when you do. Enlarging interpolates: it invents intermediate pixels by looking at their neighbours. It cannot recover detail that was never captured, so the result is a bigger, softer version of the same image.

That warning matters because a lot of resizers quietly produce a blurry 4000-pixel file from a 500-pixel source and let the user assume something was gained. Modest enlargement — up to around 150 per cent — is often acceptable for screen use. Beyond that it usually looks worse than the original displayed larger, and no amount of resampling fixes it. Genuine upscaling needs a trained model, which is a different tool with a different cost.

Which sizes to actually use

For a full-width photo on a website, 1600 pixels wide covers most desktop layouts including higher-density screens. For a content image inside a text column, 800 to 1000. For a thumbnail, twice its display size — so 300 for a 150-pixel slot — because high-density screens address roughly two device pixels per CSS pixel.

For email attachments and chat, 1600 on the long edge keeps the file small while staying good enough to view full-screen on a phone. For printing, the arithmetic is different: 300 pixels per inch of final size, so a 4 × 6 inch print wants 1200 × 1800. Below about 150 pixels per inch, print output looks visibly soft.

For profile photos and ID uploads, read the requirement carefully — those systems often specify both pixel dimensions and a maximum file size, and they are the case where you will need this tool and a compressor in sequence.

The resampling, the format, and the privacy

Scaling is done by a WebAssembly build of a proper resampling filter rather than by drawing into a canvas at a different size. That difference is visible when shrinking by a large factor: a naive canvas scale drops pixels and produces jagged, aliased edges, while a real filter averages the source region and keeps fine detail smooth.

Output can stay in the original format or switch to JPG, PNG or WebP. Note that a PNG photo resized and kept as PNG is still a PNG photo — lossless compression is a poor fit for photographic content, so choosing JPG usually shrinks the file dramatically on top of the pixel reduction. The quality slider disappears for PNG because PNG is lossless and has nothing for it to control.

Everything runs in this browser tab, which for images is not a small point: photographs carry EXIF metadata that can include GPS coordinates, camera serial numbers and timestamps. A server-side resizer receives all of that along with the picture. Here the image is decoded, resampled and re-encoded locally — and because the re-encode writes fresh pixel data with no metadata block, the location tag does not travel with the output either.