Compress JPEG photos — re-encoded at the lowest quality that still looks right.

What happens to a JPEG

The photo is decoded, then re-encoded with mozjpeg — Mozilla's encoder, with a table tuned for perceived quality rather than peak signal-to-noise — at the lowest quality that keeps the target. Each attempt is scored with SSIMULACRA2 and the search walks down until it would go below target, so you get the smallest file that still passes.

Big photos get a slightly lower quality than small ones, because a large image is usually seen scaled down, where the difference is not visible. That is the same judgement TinyPNG makes; unlike TinyPNG, this keeps a colour profile that is not sRGB, and keeps the orientation, so a photo shot in Display P3 or held sideways survives the trip. The rest of the camera metadata is dropped.

The output is never bigger than the input: if re-encoding does not help, you get the original back.

Questions

Does it strip EXIF? Yes, except orientation and a non-sRGB colour profile. EXIF carries GPS coordinates and camera serial numbers; keeping it is rarely what you want from a compressor.

Will it fix a photo that is already compressed to death? It will not make it worse, but there is little left to remove — expect single-digit percentages.

Can I convert to WebP instead? Yes, on the WebP page — at the same automatic quality, turned upright.