Convert ARW to TIFF

Archives and records offices publish lists of accepted formats, and TIFF is on nearly all of them for good reasons: the specification is open and published, it has been stable since 1992, there is no licensing question, and uncompressed pixel data can be read by software written from the specification alone. Those are properties of the container.

They say nothing about what goes in it. A TIFF made here holds a picture that was compressed once by the camera, rendered to eight bits, with the Creative Style, the white balance and the noise reduction the camera chose already applied and unchangeable. It is a faithful, durable copy of that particular rendering — which is a legitimate thing to deposit and is not the same as depositing the photograph.

This is the Sony-specific thing worth checking before an archive deposit. Sony’s compressed raw is genuinely lossy: it applies a tone curve and stores differences between neighbouring values in a way that throws information away, rather than merely packing the same numbers more tightly. For a long stretch of Alpha history it was the only raw option available, and it is still the default on some bodies.

Lossless compressed and uncompressed options arrived later and on fewer models, and if the intention is to preserve everything the sensor recorded then which of them was selected matters a great deal. It does not change what this page produces — the embedded preview is written the same way regardless — but it changes what the ARW you are keeping alongside is actually worth, and an archive should know.

Converting consumes nothing. The ARW keeps its 14-bit sensor readings, its full highlight and shadow latitude, its white balance as a changeable parameter rather than a decision, and its complete metadata block. Everything that makes it possible to produce a different, better rendering of the same frame in ten years is in that file and in no other.

That is the argument for depositing both where the archive allows it. A TIFF is the copy anybody can open today without special software; the ARW is the copy that still contains the photograph. Deposit only the TIFF and you have committed the collection permanently to one eight-bit interpretation made by a camera’s processor in a fraction of a second.

Multiply the pixels by four and add a kilobyte. A 24-megapixel A7 III frame at 6,000 by 4,000 gives 96 MB. A 61-megapixel A7R IV or A7R V at 9,504 by 6,336 gives 241 MB. A 50-megapixel A1 at 8,640 by 5,760 gives 199 MB. A 12-megapixel A7S III at 4,240 by 2,832 gives a comparatively modest 48 MB.

A deposit of five hundred A7R V frames is therefore about 120 GB, which is a storage conversation rather than an upload. Archives generally have ingest limits per file and per batch, and the sensible order of operations is to ask what they are, convert ten frames, confirm the ingest accepts them, and only then run the set.

No compression, one strip holding the whole image, photometric RGB, eight bits per sample, four samples per pixel with the fourth declared as alpha, and big-endian byte order as the original specification defined it. There are no tiles, no predictors and no planar configuration to negotiate.

That plainness is the property an archive actually wants. A file that uses only the baseline can be decoded by anyone with the specification and a text editor’s worth of patience, twenty years after every piece of software that wrote it has stopped running. Every optional feature added to a TIFF is another thing a future reader has to have implemented, and none of them are used here.

TIFF supports up to 32 bits per channel and is routinely used at 16 for exactly the preservation reasons this page is about. What this conversion writes is eight, because the embedded preview inside an ARW is an eight-bit JPEG and there is nothing deeper to write.

The gap is not academic for an archive. Eight bits per channel gives 256 levels per colour, which is enough to display the picture and not enough to survive significant reinterpretation — a future user brightening the shadows of a deposited TIFF will find banding where a 16-bit development of the ARW would have had room. Record in the catalogue that the file is an eight-bit derivative, so nobody in 2040 mistakes it for a master.

No ICC profile is written, so a reader has nothing to go on and assumes sRGB. Sony bodies offer sRGB and Adobe RGB, and a camera set to Adobe RGB writes an Adobe RGB preview whose numbers will be misread as sRGB — the visible result being saturated colours that come back tamer than they should.

For a deposit this is worth documenting rather than merely avoiding. Note the camera’s colour space setting in the record alongside the file, because a TIFF with no profile is ambiguous by construction and the ambiguity is only resolvable from information that is not in the file. If the collection has any colour-critical material in it, that material should be developed from the ARW in colour-managed software instead.

Nothing from the metadata block reaches the TIFF: no capture date, no camera or lens, no exposure settings, no copyright field, no GPS coordinate. Both the source and the target format have room for all of it, and the decode-and-encode path in between has no way to carry it.

This is less of a problem for an archive than for anyone else, because an archive’s whole method is to hold the description separately from the object. Extract the EXIF from the ARW files, record it in the catalogue with the accession, and the TIFF’s emptiness stops mattering. What must not happen is a deposit where the only record of when a photograph was taken was inside a file that no longer has it.

The converter takes the largest embedded JPEG above 640 by 480 and reports a clear failure rather than handing over a thumbnail. Files straight from a Sony always have a full-resolution preview, so a failure means the file has been through something else.

In an archive context that is a signal worth acting on. ARW files rewritten by third-party software, recovered from a damaged card, or truncated during a transfer can lose the preview while the sensor data remains readable. If a deposit batch produces failures, treat them as a provenance question about those files rather than a converter problem, and go back to the original card copy if one survives.

The ARW is read from disk by the page, scanned for its preview, decoded and written out as a TIFF, all locally. No request carries the image — the network tab shows it and a disconnected machine proves it — and there is no account, no queue and no retention policy to read.

For material under embargo, under licence, or covered by a research ethics condition, that is not a convenience but a requirement. A great many free converters upload, and a deposit workflow that routed restricted photographs through an unnamed third party would be a breach whether or not anything went wrong.