Convert JXL to PDF

This conversion is almost never about the picture. It is about a system on the other side: a claims portal, a university submission page, an expenses tool, a records office, a printer’s upload form. Those systems accept PDF because PDF is the format that is guaranteed to look the same on their screen as on yours, and many of them accept nothing else.

JPEG XL is at the far end of that spectrum. The registry files it as niche with partial browser support, Chrome removed its decoder in 2023, and no upload form anywhere is going to list .jxl among its accepted types. Wrapping the picture in a PDF sidesteps the entire question of whether the recipient can decode the format, because what they receive is a document with a picture already drawn on it.

PDF does not hold arbitrary image formats. The writer used here embeds JPEG and PNG directly, and a JPEG or a PNG dropped on this site therefore goes into the document byte for byte, untouched and instant. A JPEG XL cannot take that path — nothing in the PDF specification knows what it is — so it is decoded to pixels and re-encoded as a PNG on the way in.

The consequence is worth knowing before you look at the result: nothing is lost, and the file gets a great deal bigger. PNG stores every pixel exactly, which on a photograph means storing every grain of noise, and JPEG XL exists precisely because it does not have to. A 3 MB JXL photograph can become a PDF several times that size. If the document only has to be looked at, converting the JXL to JPG first and then that JPG to PDF produces a far smaller file by the same route the site’s own JPEG path takes.

Fit to image is the default and it means what it says: the page becomes exactly the size of the picture, so there is no white border, no letterbox and no guessing at a paper size for something that will never be printed. Most images converted to PDF are going into an email or an upload field, and there a page shaped like the picture is the better result.

The mechanism has one quirk worth knowing. PDF measures in points of 1/72 inch, and fit-to-image maps one pixel to one point, so a 4,032-pixel-wide photograph produces a page about 56 inches across. On screen this is invisible — the viewer scales it to the window — but a print dialog will show you a page the size of a door. That is the reason A4 and US Letter are there, and the reason to choose one the moment a printer is involved.

The margin control adds space around the picture. On a fixed paper size that space comes out of the image, which is scaled down to fit inside it; on fit-to-image the page grows instead, since there is no fixed boundary for the margin to eat into. A margin of zero is the default and is right for an upload, where a border is wasted bytes and nothing else.

Orientation applies to A4 and US Letter and is ignored by fit-to-image, which takes its shape from the picture and therefore cannot be in the wrong one. Automatic is the sensible default: it leaves the paper portrait, which is what a form expects. Choose landscape deliberately when the image is wide and the page is going to be printed, otherwise a panorama lands as a thin strip across the middle of an upright sheet with most of the paper unused. The picture is always centred in whatever room the margin leaves, so there is no way to place it in a corner from here — that is a layout decision, and it belongs in a document editor rather than in a converter.

The picture is scaled down to fit inside the margins and never scaled up. If the JPEG XL is 400 pixels wide and the page is A4, it is drawn at its own size in the middle of the sheet rather than stretched across it — because enlarging a small image onto a large page turns a crisp logo into a blurry one and there is no way to make that look deliberate.

So a small source produces a small picture on a large page, and that is the correct outcome rather than a bug. If you actually want the image to fill the sheet, the thing to change is the image: enlarge it in an editor where you can choose the interpolation and see the result, or go back to whatever produced the JPEG XL and export it larger. A PDF writer is the wrong place to make that decision blindly.

A PDF made this way contains one thing: a picture drawn on a page. If the JPEG XL is a scanned letter, a photographed receipt or a screenshot of an invoice, the words in it are pixels and stay pixels. Nothing in the document can be selected, copied, searched or read out by a screen reader.

That is often perfectly acceptable — a human is going to look at it, and the picture is the evidence. It fails when the recipient’s system searches submitted documents, or when you will need to find this file again by its contents in two years. It also fails accessibility outright: a screen reader announces an image and nothing else, so a form submitted this way is unreadable to somebody who cannot see it. In that case run the image through text recognition to get the words as text, and keep the PDF for what it is good at, which is looking exactly the same everywhere.

Each conversion produces a single-page PDF containing a single image. Dropping twenty JPEG XL files gives you twenty PDFs, not one document with twenty pages, and there is no ordering step because there is no sequence to order.

If what you need is one document, that is a merge and it belongs to a PDF tool rather than an image converter — the ordering, the page numbering and the rotation of individual pages are decisions this page has no way to ask you about. Converting first and merging afterwards is two steps and gets a better result than a converter guessing at the sequence of files from their names.

If the PDF is going on paper, choose A4 or US Letter and set the orientation to match the picture. That produces a document a print dialog understands immediately, with the image centred and scaled to fit inside whatever margin you asked for, and no surprise at the printer.

A margin of a few units is worth adding for print even though zero is right for uploads, because most desktop printers cannot put ink right to the edge of the sheet and will silently shrink a full-bleed page to fit their own hardware margin. Deciding the border yourself is better than letting the driver decide it, and it means the printed result matches the preview.

The JPEG XL decoder and the PDF writer are WebAssembly and JavaScript loaded into this tab on demand. The image is read from your disk, decoded in memory, drawn into a document and handed back as a download. No request carrying the picture or the finished PDF is made, so there is no server copy and no queue.

This matters more for this pair than for most, because of what people convert to PDF. The images that end up in a claims portal or a records system are payslips, medical letters, passports, invoices and signed forms — the exact category of document where “we delete it after an hour” is a promise you would rather not need. Here there is nothing to delete, and the network tab shows it.

If the recipient is a person who will look at the picture and nothing more, a PDF adds a layer of packaging and a lot of bytes for no benefit. An image format they can open is smaller, faster and easier to view on a phone, and the whole question is which image format that is rather than whether to make a document.

PDF earns its place when the file must survive a process: a submission that validates the type, a printer that needs a paper size, an archive that expects documents rather than pictures, a signature workflow. Where none of those apply, the JPEG XL is better kept as an image — and the conversion worth doing is to a format that opens everywhere rather than to a document that wraps one.