Picture Batch RAF to JPG
RAF to JPG
Fujifilm's X-Trans sensors do not use the colour filter array everyone else uses, and that single fact is why RAF files have a reputation for converting badly. Rather than a repeating two-by-two pattern, X-Trans uses a six-by-six one, and a converter that treats it as an ordinary Bayer array produces mushy detail and maze-like artefacts in fine texture. This page uses the demosaic that was written for it.
Drop RAF images here
RAF files — or any of PNG, JPEG, WebP, AVIF, BMP, ICO, TIFF, GIF or camera raw (CR2, CR3, NEF, ARW, DNG, ORF, RW2). Everything runs on this device.
Your images never leave this device. They are decoded and re-encoded by your own browser, and the security policy on every page blocks this site from sending them anywhere.
Lossy, no transparency, opens absolutely everywhere, small for photographs.
The standard range for photographs. Hard to distinguish from the original.
Images keep their original dimensions.
JPEG cannot store transparency. Nothing in the queue currently has transparent pixels, so this has no effect yet.
Exif is carried across where both formats can hold it. Nothing currently in the queue has any.
Every pixel gets its two missing colours interpolated from its neighbours, at the full sensor resolution. The slower and more detailed option.
The value your camera recorded when the shutter fired — what you saw on the back of the camera.
Camera, lens, exposure and capture time are written into the JPEG when those fields are present. The file details show whether GPS data exists, but coordinates are deliberately not written, and colour comes out as sRGB.
What happens to your image
Expect a much smaller file than a lossless source, at the cost of some detail.
| Property | Outcome |
|---|---|
| Transparency | Preserved: RAF images have no transparency to lose. |
| Image quality | Partly preserved: JPEG is lossy: the image is re-compressed and detail is discarded to make the file smaller. The quality control decides how much. |
| Colour | Partly preserved: Converted to sRGB. Decoding an image colour-manages it into the browser's own space, so the output is sRGB and is left untagged, which every viewer reads as sRGB. A wide-gamut source is reported before conversion. |
| Exif metadata | Preserved: Carried across if you leave metadata switched on, or removed completely if you switch it off. Camera, date and any location tag travel together. |
| Orientation | Preserved: Applied to the pixels, and the tag is reset to upright afterwards so nothing rotates the image a second time. |
Why X-Trans needs its own algorithm
In a conventional Bayer array every two-by-two block contains one red, one blue and two green photosites, and interpolation can rely on that regularity. X-Trans uses a larger, deliberately irregular six-by-six pattern designed to break up the repeating structure that causes moiré, which is why Fujifilm cameras can omit the anti-aliasing filter and resolve more detail.
The cost is that the interpolation is much harder. The algorithm used here works out the local direction of detail and interpolates along it rather than across it — the approach the raw-processing community settled on for X-Trans after several worse attempts — and it is several times more expensive than the Bayer equivalent.
That expense is real and worth stating plainly: a 26-megapixel X-Trans file takes several seconds to demosaic at full resolution, and noticeably longer in some browsers than others. It is arithmetic, not a stall.
Half resolution is a genuinely good option here
Half-resolution decoding uses LibRaw's reduced-resolution path, avoiding the full interpolation cost and using substantially less memory. A 26-megapixel file produces roughly a quarter as many output pixels, which is still larger than most web placements need.
Because the interpolation is what X-Trans makes expensive and half-resolution skips it entirely, this is the one format where the fast option is dramatically faster rather than marginally so. For a batch headed to a website it is the setting to use.
Fujifilm colour, without Fujifilm film simulations
Fujifilm's film simulations — Provia, Velvia, Classic Chrome, Acros — are applied by the camera when it makes a JPEG, and they are a large part of why people buy the cameras. They are not in the raw data. A converted RAF is neutral, and it will not look like the Classic Chrome image on the back of the camera.
What does come across is the measured colour profile for your sensor, so the colours are accurate rather than approximate, and the white balance the camera recorded. If the film simulation is what you were after, Fujifilm's own software and several raw editors can apply it; a converter cannot invent it.
If something goes wrong
Full-resolution conversion takes several seconds a file.
The X-Trans demosaic, as described above. This is the expensive one.
What to do: Use half resolution for anything that does not need every pixel. Otherwise start the batch and leave it — progress is per-file and real, and any file can be cancelled.
Fine detail looks slightly different from Fujifilm's own converter.
Different X-Trans demosaicing algorithms make different trade-offs between resolution, colour artefacts and noise. There is no single right answer.
What to do: Nothing to fix. If you need Fujifilm's exact rendering, use Fujifilm's software; this is a well-regarded independent implementation rather than a copy of theirs.
A file from an EXR-sensor Fujifilm compact behaves differently.
Fujifilm's older EXR and SuperCCD sensors use a rotated array rather than X-Trans, and produce images with unusual proportions.
What to do: Support varies by model. When the decoder recognises the file, the queue reports the sensor layout and output dimensions; otherwise it fails explicitly.
Questions
Do compressed and uncompressed RAF files both work?
The decoder supports uncompressed and several compressed RAF variants for known bodies. Exact support depends on the camera and compression mode; unsupported files fail explicitly. Information discarded by a lossy mode cannot be restored.
Will I lose the film simulation?
It was never in the raw file. The simulation is applied when the camera writes a JPEG. If you shot raw plus JPEG, the camera's JPEG already has it.
Why is the image bigger than my camera says it is?
X-Trans sensors are often reported at a slightly larger pixel count than the marketing figure, and the decoder returns the full usable area rather than the camera's crop. The queue shows the exact dimensions before you convert.