Convert ORF to JPG

Most raw formats hold a straightforward capture and a preview of it. Olympus and OM System bodies frequently hold something the camera assembled: High Res Shot combines several sensor-shifted exposures into a much higher-resolution image, Live Composite builds a long exposure that only accumulates new light, in-camera focus stacking merges a bracket into one sharp frame, and Live ND simulates a neutral density filter.

In every one of those the interesting picture is the computation, and the computation happened inside the camera. The embedded preview is the finished result. Third-party raw developers frequently cannot reproduce these modes at all, which turns the usual argument on its head — extraction is not the lesser path here, it is often the only one that gives you what you shot.

ORF is a TIFF-derived container Olympus has written since 2002, carrying Micro Four Thirds sensor readings — 12 bits per channel, according to the format record behind this site, rather than the 14 that the larger-sensor formats use. Alongside them the camera writes a complete JPEG at the moment of capture.

The converter scans the file for embedded JPEGs, keeps the largest above 640 by 480 and returns it. No demosaicing happens, so the process is fast enough to run through a card of files on a phone, and the picture you receive is the camera’s rendering with its Picture Mode applied.

Latitude scales with bit depth and with sensor size, and Micro Four Thirds is smaller than APS-C or full frame. A 12-bit ORF has real headroom, and less of it than a 14-bit file off a full-frame body — which means the gap between "the camera’s rendering" and "what a careful development could recover" is narrower on this system than on most.

That is not an argument for throwing away the ORF; a blown highlight is still recoverable from raw and still not recoverable from an 8-bit preview. It is an argument for not feeling that extraction is a compromise on ordinary frames. For a correctly exposed picture, the camera’s JPEG is very close to what you would have arrived at anyway.

The system’s whole appeal is weight, and the people who chose it are disproportionately away from their main machine: on a mountain, in a hide, three weeks into a trip with a tablet and a card reader. That is the awkward moment, because ORF is invisible to browsers, needs a codec on Windows, and OM Workspace is not going onto an iPad.

This runs in the browser tab with nothing installed, so the tablet and the card reader are a complete filing setup. It also works with the connection off, since nothing is being sent anywhere — useful on a hotel network that charges by the megabyte.

The preview is passed through without a re-encode, so everything the camera wrote into it could come with it: lens and focal length, aperture, shutter speed, ISO, capture time, and GPS coordinates where the body or a paired phone supplied them. Whether it does is a setting. "Remove metadata" is on by default, so the EXIF is edited out before you get the file, and what survives regardless is the ICC colour profile.

For a wildlife or landscape photographer that default is the useful one — nest sites and specific spots get shared far too readily by accident, and a JPG posted from a trip is exactly how it happens. When you do want the data, for your own records or for a submission that asks for capture details, turn the option off and the block arrives intact. Setting a maximum width is no longer the way to strip metadata: it only ever did so by forcing a re-encode, and that spends quality for nothing.

The Picture Mode is baked in. A frame shot in Monochrome comes out monochrome and the colour is not retrievable from the JPG, though it is still sitting in the ORF. Auto white balance that misjudged a scene under snow or forest canopy is likewise fixed in place.

Highlight recovery is the clearest limit. A bright sky rendered as flat white in the preview usually still holds detail in the sensor data, because the preview was made with a tone curve that clipped it. If a frame is worth the effort, keep the ORF and open it in OM Workspace, Lightroom or darktable — all three read it.

Anything below 640 by 480 is rejected rather than returned, because a thumbnail dressed up as a conversion is worse than a clear failure. That threshold is almost never a problem on files straight from a camera.

It becomes one on files that took a detour: recovered by an undelete tool, truncated by a card removed mid-write, or written by third-party software that did not include a preview. The original card copy is the first thing to try; failing that, the file needs a decoder that reads the sensor data.

Drop the whole selection and take the result as a ZIP. Each file is scanned separately with its own progress, and because a JPG target at the default settings involves no re-encoding at all, the limit is how fast your device can read the files.

The free tier accepts up to 100 MB per file, which no ordinary ORF approaches — a 20-megapixel frame is around 20 MB — and which even a High Res Shot file stays inside.

The conversion is local: the ORF files are read from your card or disk, scanned in memory, and the JPGs written back. Nothing is transmitted at any point, which matters on a shared hotel connection as much as it does for privacy.

You can verify it instead of trusting it. Open the developer tools and watch the network tab during a conversion, or put the device into aeroplane mode and convert anyway.