Image tools

Image tools are the ones most worth being careful about, because "upload your photo to resize it" means giving a stranger the original file and whatever metadata it carries. These two do the work with the Canvas API inside your browser: the file is read from disk by the page, processed in memory, and handed back as a download. Nothing is transmitted, which you can verify by going offline before you start.

Resize when a dimension is the problem — an image that is 4000px wide on a page that displays it at 800px wastes bandwidth and hurts Core Web Vitals for every visitor. Convert when the format is the problem: WebP is typically 25–35% smaller than a comparable JPEG, PNG is the right answer for screenshots and anything needing transparency, and JPEG remains the safe choice for photographs going somewhere old.

Dimensions first, then format

If both are wrong, resize first and convert second. Encoding a large image and then shrinking it throws away quality you already paid for, and a JPEG that has been through two rounds of lossy compression shows blocking around edges that a single pass would not have produced. The order matters even more for WebP and AVIF, whose encoders make decisions about which detail to keep based on the pixels they are given; give them the final dimensions and they will spend their bits where it counts.

A useful rule for the web: export at twice the largest size the image will be displayed at, and no larger. A hero image shown at 1200px on a wide screen wants a 2400px source for high-density displays and gains nothing from 4000px. For thumbnails and avatars the multiplier is the same, which is why a 160px avatar exported at 320px looks sharp and a 1000px one is just slow.

Choosing between JPEG, PNG and WebP

The format decision is about what is in the image, not what it is for. Photographs and anything with gradients compress well with lossy encoders — JPEG or WebP — because the eye does not notice small errors in smoothly varying colour. Screenshots, diagrams, logos and text do not: JPEG smears crisp edges, so those want PNG or lossless WebP. Transparency rules out baseline JPEG entirely. WebP handles all of these cases and is smaller in nearly all of them, and every current browser decodes it; the remaining reason to produce JPEG or PNG is that the image is going somewhere that is not a browser — an email client, a document, an old CMS that rejects anything else.

Quality settings on lossy formats are not a percentage of anything measurable, so treat them as a dial rather than a specification. Between 75 and 85 is where most photographs stop visibly improving; the converter here shows the original next to the encoded result and the size of each, because the only judge that matters is your own eye on the actual image.

What the browser does and does not strip

Drawing an image through a canvas produces new pixels and a new file, which means EXIF metadata — camera model, GPS coordinates, timestamps — does not survive the trip. For anything you are about to publish that is usually what you want. Be aware that it also removes colour-profile information, so a photograph edited in a wide-gamut space may look slightly flatter after conversion; if colour accuracy is the point, do the conversion in the tool that produced the image instead.