Convert RAF to JPG

A film simulation is a rendering, not a recording. The sensor captures light; the camera’s processor turns that into Classic Chrome or Acros by applying Fujifilm’s own colour science, tone curves, grain and colour chrome effects. Open the RAF in third-party software and none of that has happened yet — the software has to demosaic the file itself and apply its own approximation of the simulation.

The JPEG the camera embedded at capture has all of it applied already. That is why extracting it produces the picture you framed and approved, and why on Fujifilm specifically this route is not the lesser option. If you bought the camera for how its files look, the camera is the thing that made them look that way.

Almost every other sensor in this bundle uses a Bayer colour filter array: a repeating two-by-two pattern of red, green and blue. Decades of demosaicing research assumes it. Fujifilm’s X-Trans uses a six-by-six pattern with a less regular distribution, chosen because it suppresses moiré well enough to omit the optical low-pass filter and gain a little sharpness.

The cost is that generic demosaicing does not apply. Converters have had to write X-Trans-specific code, and the results have been argued over for years — smeared foliage, a wormy texture in fine detail, and painterly rendering of grass and leaves are complaints Fujifilm shooters have levelled at various versions of various converters. Extraction sidesteps the whole argument, because the camera already demosaiced the file with the algorithm designed for its own sensor.

CR2, NEF, ARW, ORF and DNG are all TIFF derivatives, so a tool that can walk TIFF structure can usually find its way around them. RAF is not — it is Fujifilm’s own container, identifiable by the literal string FUJIFILM at the start of the file, with its own arrangement inside.

This converter handles it for an unglamorous reason: rather than parsing container structures it scans the whole file for JPEG start markers, validates each candidate and keeps the largest above 640 by 480. That works identically on a TIFF-based CR2 and on an RAF, which is worth more here than an elegant parser that handles five formats out of seven.

Exposure. The extracted JPEG is 8-bit with the tone curve applied, so a clipped highlight is gone and a shadow lifted two stops falls apart. The RAF holds 14-bit sensor readings and both of those are recoverable from it in Lightroom, Capture One or darktable.

The other case is wanting a different simulation than the one you shot. Fujifilm’s own X RAW Studio and Capture One’s Fujifilm profiles can apply a different look to the raw data — X RAW Studio notably does it using the camera’s own processor, which is the only way to get an exact result. If you shot Provia and wish it were Acros, that is where to go; this page will only ever give you Provia.

A lot of Fujifilm shooters set the camera to record a JPEG alongside every raw for exactly this reason: the camera JPEG is the deliverable and the raw is insurance. It is a good habit and it costs a little card space.

This page is what you need on the day you forgot, or on the card that was set to raw only, or when somebody hands you RAF files and no JPEGs. The picture is in the file; it was always in the file. Nothing is lost by not having enabled the setting.

Everything. Extraction reads the file and writes a new one; the RAF is untouched, with its full sensor data, and can be developed later if a frame turns out to matter more than it seemed.

That is worth stating because the natural assumption with a converter is that something has been consumed or downgraded. Here the raw file is a negative that stays in the drawer, and the JPG is a print made from the preview the camera already produced.

Because the JPEG is copied out rather than re-encoded, the EXIF block can survive intact: lens, focal length, aperture, shutter speed, ISO, capture time and any GPS coordinate supplied by a paired phone. The film simulation in use is generally recorded there too, which is useful when you are trying to remember what produced a look you liked. Whether it survives is up to you — "Remove metadata" is on by default, so unless you turn it off the JPG arrives without any of it.

Leave the default alone for pictures going somewhere public and the location does not travel with them. Turn it off when the simulation tag is the thing you are after, or when the set is going to somebody who needs the capture data. What is never dropped is the ICC colour profile, and on a wide-gamut file that is the difference between the colours the camera intended and a washed-out version of them. Setting a maximum width used to be the way to strip the rest; it works only by re-encoding the picture, and there is no longer any reason to spend that.

Drop the whole selection and take the finished set as a ZIP. For a JPG target at the default settings there is no decoding and no encoding — the preview is located and copied — so the work is bounded by how fast the files can be read.

The free tier accepts input up to 100 MB per file. A 40-megapixel X-T5 RAF is comfortably inside that. An uncompressed frame from a 102-megapixel GFX body is not — it is around 180 MB at 14 bits — so those bodies have to be set to compressed RAF, which is the sensible setting on them anyway.

The conversion runs in the browser: the RAF is read locally, scanned locally and the JPG written locally. Nothing is transmitted at any point, which is the difference between a converter you can use on a paid job and one you cannot.

It also means it works on a machine where you cannot install X RAW Studio or Capture One — a work laptop, a borrowed computer, a tablet with a card reader — and it works with the network disconnected, because there is nothing for the network to do.