Compress JPEG XL

JPEG XL came out of a competition to replace JPEG properly, and its priorities show: it encodes far faster than AVIF at similar sizes, supports images beyond what a screen can display, and can repack an existing JPEG losslessly at around a fifth smaller — the original recoverable byte for byte.

That last capability is unique and is what makes it interesting for archives. It is not what this page does, which is ordinary lossy re-encoding, but it is worth knowing that a JPEG collection can be shrunk without any loss at all by a tool built for it.

Browser support is the unresolved part. Safari ships it, Chrome removed it after shipping it behind a flag, and Firefox has not enabled it. Photoshop and several professional tools read it; most consumer software does not.

For a website, that rules it out — you would be serving a format most visitors cannot display. For a personal archive on your own disk, where you control what opens the files, the calculation is different and JXL is a reasonable bet.

This page re-compresses a file that is already JPEG XL, which means putting an image through the codec a second time. If the source was encoded lossily, that second pass compounds the first — the encoder is working from an image that has already had detail removed and removes more.

Where a real saving is available is at the first encode, from an original. If you still hold the JPEG or the raw file that produced this JXL, going back to it and encoding once is better than squeezing what you have. That is true of every lossy format and it is easy to forget when a tool offers a slider.

JPEG XL has a genuinely lossless mode, and a file encoded that way loses nothing however often it is re-encoded. If the file you are holding is lossless JXL, re-compressing at a higher effort setting is free of quality cost and the only question is how long it takes.

Nothing in the extension says which you have. A rough guide is size against dimensions: lossless JXL of a photograph lands in the same territory as a PNG, while lossy JXL is a fraction of that. If the file is unexpectedly large, that is usually the explanation rather than a problem to compress away.

JPEG XL exposes an effort level that decides how hard the encoder searches, and the range is wide. Low effort encodes quickly and leaves size on the table; high effort can take many times longer for a saving in the region of ten to twenty per cent.

For a file written once and read many times that is worth paying. For a batch being processed on a laptop it usually is not, and the middle of the range is where most of the benefit already sits. Decoding speed is unaffected by the choice, which is the right way round.

Safari on macOS and iOS displays JXL, Adobe applications read it, and a growing set of image tools support it. Chrome does not — support was removed in 2023 — and Firefox has it behind a flag.

So a compressed JXL is a storage decision rather than a delivery one. It is a reasonable format for an archive on your own disk, where you control what opens the files, and a poor one for a public web page, where a large share of visitors would see nothing at all.

The interesting question about a JPEG XL file is almost never how much smaller it gets than itself. It is whether the format earns its place against the JPEG the picture would otherwise be, and the honest answer is usually a saving in the region of twenty to thirty per cent at a visually equal setting.

That is a real number and it is not a large enough one to justify a format nothing can open. Where the file stays in an archive that you control and will still control in ten years, the saving compounds across a collection. Where it is going anywhere else, the saving is spent on a file the recipient cannot see.

JPEG XL buys its efficiency with a more elaborate decoder, and the effort setting that made the file small also makes it slower to open. On a large photograph the difference against JPEG is noticeable rather than theoretical, and it lands on every viewer every time.

For an archive that is the right trade — files are written once and read rarely. For anything that will be opened repeatedly, or opened on a modest machine, it is a cost worth weighing against a saving that was only a fifth of the file to begin with.