How this tool works
The image is decoded and re-encoded in your browser at the quality you choose, and the size you see is measured from the finished file rather than estimated. Dimensions are never changed: a 4000-pixel photograph comes back 4000 pixels wide, just encoded more efficiently.
A JPEG source stays JPEG. A PNG or WebP source is encoded to WebP so that transparency survives, because JPEG cannot store an alpha channel and re-encoding a transparent PNG as JPEG fills those areas with solid colour. Where a browser cannot write WebP, the image is flattened onto white first rather than left with black where the transparency was. Quality is held between 10 and 90.
Getting a real saving
- Resizing saves more than quality does: a photograph displayed 800 pixels wide gains far more from being 800 pixels wide than from any quality setting. Resize first, then compress.
- 70 to 80 suits photographs: the file drops sharply and the artefacts stay invisible at normal viewing size. Reserve 85 and above for images where fine detail carries the point.
- Screenshots and line art are the exception: flat colour and sharp text are what lossy encoding handles worst. A screenshot often compresses better as an untouched PNG than as a lossy WebP.
- A larger result means it was already optimised: when re-encoding would add bytes you are told so and given nothing to download, rather than handed a bigger file labelled compressed.