What "Compressing" an Image Actually Means
Most everyday image compression isn't really about removing detail you'd notice — it's about removing detail you wouldn't. JPEG and WebP both use lossy compression algorithms that discard information the human eye is bad at detecting, like subtle color gradients in areas with little contrast, while keeping the sharp edges and detail that actually register as "quality." That's why a well-compressed photo at 80% quality can look nearly identical to the original while being a fraction of the file size — the missing 20% was mostly information you were never going to see anyway.
Why the Upload Step Is the Part That Shouldn't Exist
Compressing an image is a purely mechanical transformation — decode the pixels, re-encode them with a different quality setting or at different dimensions, done. None of that requires a server. Every step happens the same way whether it runs on a remote machine or the one already displaying the image, which means uploading a photo to compress it is pure overhead: the time spent uploading, the privacy exposure of a photo (which may contain identifying details, documents, or private content) sitting on a third-party server, and the trust required in whatever that server does with it afterward. A browser's Canvas API can decode an image, redraw it at a new size, and re-encode it to a compressed format entirely in memory — the same operation, without ever leaving the device.
Choosing a Format: JPEG vs WebP vs PNG
The three common output formats solve different problems:
- JPEG — the most universally supported lossy format, best for photos with lots of color variation. No transparency support.
- WebP — generally produces smaller files than JPEG at equivalent visual quality, and supports transparency, but has slightly narrower legacy compatibility.
- PNG — lossless, meaning zero quality loss no matter how many times it's re-saved, but files are significantly larger. The right choice when the image has hard edges (screenshots, logos, text) or needs transparency and can't tolerate any compression artifacts.
Resizing Matters as Much as Quality
Quality sliders get most of the attention, but pixel dimensions often matter more for file size. An image resized to half its width and height has roughly a quarter of the total pixels to encode — a far bigger size reduction than any quality slider adjustment alone. If an image is going to be displayed at 800px wide on a website, shipping a 4000px-wide original wastes bandwidth regardless of how aggressively it's compressed at that resolution. Resizing to the dimensions the image will actually be displayed at, then compressing, gets the best result from both steps combined.
When Compression Isn't Worth It
Not every image benefits from lossy compression. Screenshots of text, technical diagrams, and images that will be re-edited or re-compressed repeatedly (each lossy re-save compounds quality loss) are better kept as PNG. The general rule: compress aggressively for anything meant to be viewed once and forgotten — email attachments, web images, social posts — and stay lossless for anything that's a working file or has to stay perfectly crisp.