Image tools

Two tools, one decision

Compressing and converting overlap enough that it is worth being clear about which you need. Compressing keeps the format and trades visual detail for a smaller file. Converting changes the format, which usually changes the size as a side effect but is really about compatibility — getting a HEIC photo into something a website will accept, or turning a screenshot into a format that keeps text edges sharp. If your problem is an upload limit, start with compression. If it is a file a system refuses to open, start with conversion.

Lossy and lossless are not interchangeable

JPEG and, in its usual mode, WebP are lossy: they discard information that is hard to see, and the discarded data does not come back. PNG is lossless: it reduces size by describing the same pixels more efficiently, so the image is bit-for-bit identical when decoded. This distinction decides which format suits which image. Photographs, which are full of gradual variation, compress dramatically as JPEG with little visible cost. Screenshots, diagrams, and anything with text or flat colour do badly as JPEG — the algorithm smears the sharp edges it was never designed for — and often come out smaller as PNG anyway.

This produces the result that surprises people most: converting an optimised PNG screenshot to JPEG at a high quality setting can make the file larger than the original, while also making it look worse. The numbers quoted in the tool pages here are measured in a real browser and re-checked by an automated test rather than estimated, so they reflect what the encoder actually does.

Quality settings have a useful range and a wasteful one

JPEG and WebP quality is not a percentage of anything meaningful, and the top of the scale is where the trade stops being worthwhile. Moving from 0.95 to 0.85 typically removes a large share of the file size while leaving differences that are hard to see outside a direct comparison. Above roughly 0.9 you are mostly spending bytes to preserve detail no viewer will notice. The reliable approach is to step the quality down until the image visibly degrades, then go back one step.

Resizing beats compressing

The single most effective way to shrink an image is to reduce its dimensions, because file size scales with pixel count rather than with the quality slider. A photo from a modern phone is several thousand pixels wide; displayed in an email, a document, or a normal web page, most of those pixels are discarded at render time anyway. Halving both dimensions removes three quarters of the pixels before any compression is applied. Reach for the quality setting after the dimensions are right, not instead of.

Re-encoding accumulates

Every lossy save discards information relative to what it was given, so a file that has been compressed, edited, and compressed again carries the losses of both passes. The damage is not reversible by saving at a higher quality afterwards. Keep an original and work from it each time, rather than repeatedly processing the processed copy.

Your images stay on your device

Both tools decode, resize, and re-encode in your browser using the same image pipeline the browser uses to display pictures. Nothing is uploaded, which matters more here than in most categories: photos carry location data, screenshots carry whatever was on screen, and scanned documents carry everything on them. An automated test loads real files into both tools and fails the build if any network request carries data out. Note that re-encoding also strips embedded metadata such as EXIF location — usually desirable, occasionally not, so keep the original if the capture data matters.