Convert JXL to PNG

The registry lists exactly two pieces of software for JPEG XL: GIMP and ImageMagick. That list is short because it is accurate. A format standardised in 2021 that most browsers declined to ship has not made it into the import dialogs of the tools people actually have open — Photoshop needs a plugin, Paint.NET and Canva have nothing, and Figma will not place the file at all.

PNG’s list is the opposite problem: there is no image tool worth naming that cannot open one. It has been a W3C Recommendation since 1996 and ISO/IEC 15948 since 2004, it is lossless, it carries transparency, and every editor on every platform reads it without a plugin, a flag or a codec pack. That combination is why this conversion exists — not because PNG is better, but because it is the format the software on your machine will accept.

There is no quality slider on this pair, and its absence is the point. PNG stores every pixel of the decoded image exactly, so the file you open in the editor is the picture the JPEG XL decoded to, pixel for pixel, with no second round of compression artefacts layered on top.

What it cannot do is undo what has already happened. JPEG XL has both a lossless and a lossy mode, and if your file used the lossy one, the PNG preserves that result faithfully rather than repairing it. Lossless here means this step adds nothing; it does not mean the picture is a pristine original. That distinction matters most when the JXL was itself repacked from an old JPEG, in which case you are looking at that JPEG’s artefacts stored perfectly.

Both formats carry a full alpha channel, so nothing is flattened and no background colour is asked for. A cut-out product shot, a logo with nothing behind it, a soft drop shadow or an icon with anti-aliased edges arrives in the editor exactly as it left — with the partial transparency at the edges preserved rather than fringed onto white.

This is the single clearest reason to choose PNG over JPG for the same file. JPG has no alpha channel, so the same picture converted that way gets filled with a background colour before it is encoded, and a cut-out flattened onto white is only correct until somebody puts it on a coloured background. If you intend to composite the image, PNG is not a preference here, it is the requirement.

JPEG XL can store up to 32 bits per channel and a wide colour gamut; PNG tops out at 16 and is often written at eight. The pipeline here decodes to eight-bit RGBA and encodes from that, so the PNG you get is eight bits per channel whatever the source held.

For most of what people convert this changes nothing, because the source was eight bits already — screenshots, line art, web graphics and ordinary camera JPEGs all live there. It matters for a wide-gamut or HDR file where the extra precision was preventing visible banding in a gradient. If you are editing that kind of file seriously, the JPEG XL is the master and GIMP, which does read it, is the tool that will not throw the depth away.

Large files are PNG’s listed weakness and this conversion walks straight into it. PNG compresses by predicting each pixel from its neighbours and packing the differences losslessly, which is superb on flat colour, screenshots and line art — areas where neighbouring pixels genuinely are alike — and poor on photographs, where every grain of sensor noise has to be stored exactly because none of it can be discarded.

JPEG XL was built three decades later and has a lossless mode of its own that beats PNG at PNG’s own job, on top of a lossy mode PNG does not have. So the increase is expected rather than a fault, and its size depends entirely on the picture: a flat illustration may barely change, while a 12-megapixel photograph can multiply. The variable is how much of the image is genuinely flat, which is also why a screenshot and a portrait of the same pixel dimensions can differ by an order of magnitude after conversion. Convert one file and look at both numbers before you convert four hundred, and check that the destination — a shared drive, a version-controlled repository, an email attachment limit — can take the result.

The sequence that works is: keep the JPEG XL, export a PNG, do the editing, then save the finished result in whatever the finished result should be. The PNG exists to get pixels into a program, and once that program has them it has no further use for the file.

Deleting the JXL afterwards is the one move to avoid. It is smaller, it holds at least as much, and it is the only copy that still has whatever depth and metadata the original carried. A directory that has quietly turned into PNGs three times the size, with the masters thrown away, is the failure case this section exists to prevent. It happens by accident rather than by decision: somebody converts a folder, works on a few files, and the PNGs are the ones that get backed up because they are the ones the editor left open. Keeping the two sets in separate directories from the start is a cheaper fix than recovering masters that no longer exist.

A JPEG XL can hold EXIF, XMP and an ICC profile, and PNG can hold all three as well — but none of them survives the trip. Every browser-side image conversion here decodes to a raw pixel grid and encodes again, and a pixel grid has nowhere to keep a camera model, a capture date, a location or a colour profile.

The “remove metadata” switch is on by default and describes what happens regardless of its position. For a file heading into an editor that is usually irrelevant, and for a file heading onto the web it is a quiet privacy win. It is only a problem if the PNG is meant to be the archive copy, and it should not be — that is what the JPEG XL you kept is for.

PNG is listed for screenshots, line art and web use, and those are the cases where the size penalty largely disappears. A screenshot is mostly flat regions of interface colour with sharp text on top, which is exactly the input PNG’s predictor was designed for, and a diagram or a UI mock-up behaves the same way.

So the decision is easier than it looks. If the JPEG XL contains flat artwork, a screen capture or a drawing, converting it to PNG costs you very little and buys universal compatibility, and the result is a perfectly reasonable file to keep. If it contains a photograph, the PNG is a temporary means to an end and the sentence above about not deleting the original applies with full force.

A single file is the rare case. People who have JPEG XL at all usually have a directory of it, because the format arrives in bulk — from an archive export, a photo library conversion or a tool that quietly chose it as its output. Drop the whole folder and each file is converted in turn, with progress per file, and the results come back together.

Because the conversion runs in the tab rather than on a server, the batch is bounded by your processor and your memory, not by an upload queue or a daily allowance. The free ceiling is 100 MB per file. A large batch of photographs is worth watching, though — not because it will fail, but because the PNGs it produces may need several times the disk space the JPEG XL originals did.

The JPEG XL decoder and the PNG encoder are both WebAssembly modules loaded on demand into this page. The file is read from disk by the tab, decoded in memory and written back out, and no request carrying the image is made at any point. There is no account, no queue and no server copy to be deleted afterwards, because nothing was ever sent.

That is checkable rather than a claim to take on faith: open the network tab, convert a file, and watch the codecs download and the image not. It matters more than usual for design work, where the file is frequently unreleased artwork, a client’s brand asset or a screenshot of something that is not public yet.