Convert MP4 to JXL

JPEG XL came out of the JPEG committee, was standardised as ISO/IEC 18181-1 in 2022, and compresses both losslessly and lossily. The lossless mode is the reason most people are interested: it is the property that makes the format a candidate for keeping images rather than serving them.

This page does not write it. The encoder here produces lossy JPEG XL at quality 82, the same default every lossy target on this site uses, and there is no control that switches modes. Stated first rather than in a footnote, because somebody moving a still archive to JXL on the strength of its lossless reputation would otherwise get the opposite of what they came for.

Type a number of seconds and the conversion takes the frame being displayed at that moment — formally, the last frame whose own start timestamp falls at or before the number. Leave the box at 0 and you get the first frame of the video, which is the default and rarely the interesting one.

It is one frame per conversion. This is not a sequence exporter and there is no option that makes it become one; extracting fifty stills means fifty timestamps. Asking by timestamp rather than by frame number is deliberate — the frame rate of real material is 25, 30, 29.97 or whatever a phone chose mid-recording, and frame arithmetic over that is a trap.

Not much headroom. The frame arrives as eight bits per channel, with colour recorded at a fraction of the resolution of brightness, and with fine detail already removed by the video encoder wherever motion made it cheap to remove. JPEG XL’s ability to carry 32 bits per channel and wide colour has nothing to attach itself to.

That is not an argument against using it — it is an argument against expecting the format to improve the picture. A lossy JXL of a video frame is a competently compressed copy of an already-compressed image. Where the format earns its place here is size, and only size.

On one frame, nowhere worth caring about. The difference between a JXL and a JPG of the same still is small enough that nobody notices it on a single file, and both are trivial next to the video they came from.

On ten thousand frames it is a different conversation, and that is the case this page exists for: a long-running capture project, a reference archive off hours of footage, a personal collection that has been accumulating for years. That is where a consistent percentage compounds into something that changes what fits on a disk — and where the format question is worth thinking about at all.

Browser support is partial and has been contested rather than steadily improving. GIMP and ImageMagick read the format, several photo managers do not, and plenty of ordinary software will show a file it cannot open. JPEG XL is a niche format that deserves better and has not yet had it.

For an archive that is the real risk, and it is not about quality. A library you cannot open on a borrowed laptop, or hand to somebody, or drop into a document, has a practical cost every time you touch it. Keep a converter close, or keep the archive in something duller.

PNG. It is lossless, standardised since 1996, read by every browser and every editor, and it will still open in thirty years — which is the only property an archival format is really judged on. TIFF is the other answer where a print or records workflow specified it.

Neither compresses as well as lossless JPEG XL would, and neither is available on this page in a mode that loses information. That is the trade: JXL here is the small file, PNG is the exact one, and choosing between them is choosing between disk space and certainty.

No downscaling happens. A 4K video gives a 3840 by 2160 image, and a 720p one gives 1280 by 720, whatever size the player happened to be on screen when you found the moment. That is the main reason to extract a frame rather than take a screenshot of a paused player.

Nothing adds resolution the recording did not have. If the video is 720p, the still is 720p, and no format on this site or elsewhere changes that — upscaling produces a bigger file containing exactly the detail that was already there.

The file carries no EXIF, no capture date, no camera, no copyright field and no colour profile, because the pixels are read out of a decoder and everything the video knew about itself stays behind in the video.

For an archive that is a gap to fill deliberately rather than a detail. Record the source file and the timestamp somewhere — in a sidecar, in the file name, or in the metadata fields afterwards. A frame archive whose members cannot be traced back to the moment they came from loses most of its value within a year.

The decode and the encode both run in the browser tab on your own processor, so nothing carrying the video is sent anywhere and the network tab will show it. The free tier takes files up to 100 MB.

For a frame-extraction habit that matters more than for a one-off. People who pull stills out of video in quantity are usually working from source material — unreleased footage, research recordings, client work — and a workflow that ships each of those to a server to get one picture back is one you would have to explain.