The question people ask is "which format has the best quality for archiving". That is the wrong question, and answering it leads you somewhere risky.

A file format's survival has almost nothing to do with how good it is. It depends on one thing: how many pieces of software, on how many machines, can decode it without anybody having to go and find something. By that measure the safest formats you can archive in are two of the oldest and least exciting ones available, and the format holding your most valuable images is probably the most fragile thing in your library.

Longevity is a decoder problem

A file on a disk is a sequence of bytes. It only becomes a photograph when a program that understands those bytes turns them into pixels. Lose the program and the bytes are inert, no matter how carefully you stored them.

So the things that make a format durable are boring and social rather than technical:

  • A published specification. If the format is documented and freely available, someone can write a reader in the future from first principles. If it was only ever reverse-engineered, that possibility depends on someone's old code still existing.
  • No licensing obstacle. A format with patent royalties attached gets implemented in fewer places, because open-source projects and small vendors cannot pay. That thins out the decoder population before you even get to the future.
  • An enormous installed base. This is the big one. Billions of existing files force support to continue, because dropping it would break too much. Formats do not survive on merit, they survive on inertia.
  • Multiple independent implementations. If only one company's code can read it, the format's future is that company's business decision.

Notice that image quality appears nowhere on that list. A visually superior format with three implementations is a worse archival bet than a mediocre format with three thousand.

The two safest bets are the boring ones

JPEG is from 1992, it is an international standard, its patents expired long ago, and essentially every device manufactured in the last thirty years can decode it. There are trillions of JPEGs in existence. Support for JPEG will not be removed from anything, ever, because doing so would break the entire internet and every photo library on earth.

It is lossy, and people hold that against it for archiving. That is a reasonable objection to re-encoding, and a poor objection to keeping a JPEG you already have. A JPEG shot at high quality that you never re-save holds its quality indefinitely. The bytes do not decay.

TIFF is older still, dating to the mid-1980s, and it is the format the archiving and museum world settled on decades ago. It stores full-quality pixel data with no lossy compression, it carries high bit depths and colour profiles properly, and the specification has been public since forever. The US Library of Congress publishes a Recommended Formats Statement for exactly this question, and uncompressed TIFF sits at the top of its preferred list for still images, with JPEG among the acceptable ones.

TIFF's cost is size. An uncompressed 24-megapixel 16-bit TIFF runs to well over 100 MB. That is the price of the guarantee.

One genuinely technical archival argument for TIFF: heavily compressed formats are fragile in the face of storage corruption. A single flipped bit early in a JPEG's compressed stream can destroy everything after it, because the decoder loses its place. In an uncompressed TIFF, a flipped bit changes one pixel. This is not a reason to convert your library, but it is a real reason archives prefer uncompressed data for irreplaceable material.

TIFF's catch: it is a container, not a format

If you archive as TIFF, this matters. TIFF is more like a filing cabinet than a format, and what is inside can vary enormously. There is baseline TIFF, which every reader handles, and then there are dozens of extensions: LZW compression, ZIP and Deflate compression, JPEG data wrapped inside a TIFF, CMYK and Lab colour, layers, and any number of vendor-specific private tags.

A TIFF written by one program using an unusual combination can genuinely fail to open in another. "It's a TIFF" is not by itself a guarantee.

The rule for archiving is to stay boring: uncompressed, or LZW or ZIP compression, RGB or greyscale, one image per file, no layers, no proprietary extras. That version of TIFF opens everywhere and will keep doing so. If you need layers, keep the layered working file separately and treat it as a working file, not as your archive copy.

Narrow aisle in an archive store, with paper-wrapped bundles of documents and old bound ledgers packed onto metal shelving on both sides
Archives settled on TIFF for the same reason they settled on plain boxes and steel shelving: predictable beats clever when nobody knows who will be opening it. Photo: Luke McKernan (CC BY-SA 2.0).

PNG, and where it belongs

PNG is lossless, fully documented, unencumbered by patents, and supported universally. For anything with flat colour, sharp edges or transparency, it is the correct archival choice: screenshots, logos, diagrams, scanned line art, game assets, UI elements.

For photographs, it is a poor fit. PNG's compression is designed for images with large areas of identical colour, and photographic noise defeats it completely. A photograph stored as PNG is typically several times the size of the same image as a good-quality JPEG, without being meaningfully better than a lossless TIFF for the purpose. You get archival safety either way, and TIFF handles bit depth above 8 bits per channel and colour management more comfortably.

So: PNG for graphics, TIFF for photographs. The distinction is about content, not about which format is superior.

Proprietary RAW is the real risk in your library

Here is the uncomfortable part. The files you consider most valuable, the RAW originals, are the least durable things you own.

There is no RAW standard. Canon's CR3, Nikon's NEF, Sony's ARW and Fujifilm's RAF are separate, undocumented, manufacturer-specific containers, and the internal details change from camera model to camera model. Third-party software reads them because people reverse-engineered them, not because a specification was published. We go into what is actually inside these files in what a RAW file actually is.

Every criterion from the top of this article fails: no published spec, a small number of implementations, and an installed base measured in the files of one camera model rather than in the whole world.

In practice, old RAW formats have stayed readable, largely thanks to open-source work in libraries like LibRaw and dcraw that has quietly kept decades of camera models supported. That is a real and reassuring track record. It is also a track record that rests on volunteer effort and a specific set of codebases continuing to exist, which is a different kind of promise from the one JPEG makes.

The honest summary: your RAW files are probably fine, and they are the part of your archive with a single point of failure. Which is a reason to have a plan, not a reason to panic.

DNG as a partial hedge

DNG is Adobe's publicly documented RAW container. The specification is freely available, so anyone can write a reader without reverse-engineering anything. Some manufacturers write it natively, and phone raw capture on both iOS and Android produces DNG. Adobe's free DNG Converter will rewrite proprietary RAW files as DNG, with an option to embed the original inside.

It is a sensible hedge and it is not a complete answer. Conversion is another interpretation step, some camera-specific processing information does not carry across the way the manufacturer's own software would use it, and DNG's own longevity depends partly on Adobe's continued interest, even though the published spec limits that dependency considerably.

The version of this that costs you nothing in risk: convert to DNG if you want the documented container, and keep the original RAW files as well. Storage is cheaper than regret. Never convert-and-delete.

AVIF, HEIC and JPEG XL: good formats, unproven decades

The modern formats are genuinely better at compression. AVIF and JPEG XL in particular produce dramatically smaller files at a given visual quality than JPEG, and if you want the comparison for web delivery we cover it in AVIF vs WebP.

For archiving, compression efficiency is the wrong criterion, and "new" is a mark against a format rather than for it. None of these has the one property that actually matters, which is thirty years of accumulated inertia.

Their individual situations are different in ways worth knowing:

  • HEIC is on billions of iPhones, which is a serious installed base. It also carries patent licensing obligations through the HEVC pools, and that is precisely why it never became universal on the web despite being the default capture format on the most popular camera in the world. Licensing friction limits how widely decoders spread.
  • AVIF is royalty-free and backed by a broad industry alliance, which is the healthiest possible starting position. It is also young, and derived from a video codec, which ties part of its fate to decisions made about video.
  • JPEG XL is royalty-free and technically excellent, including the ability to losslessly recompress existing JPEGs. Its browser support has been an unusually public argument, with implementations added, removed and re-added. That is a reminder that a format's fate is decided by organisations, not by its merits.

None of this means avoid them. It means do not make one of them the only copy of something you cannot replace.

Formats that were going to win

This is not a hypothetical concern, and the history is short enough to check.

JPEG 2000 arrived in 2000 as the successor to JPEG. It was better in most measurable ways. Browsers never broadly adopted it, and it stalled. It did not disappear, though: it is entrenched in digital cinema distribution, medical imaging and institutional archives, and it is well supported inside those worlds. That is the actual lesson. Formats survive where they have an installed base, and die where they do not, independent of quality.

JPEG XR, which began life as Microsoft's HD Photo, was standardised and shipped in Internet Explorer. Browser support was later dropped and it is now effectively gone from general use.

Kodak Photo CD and FlashPix were both backed by major companies in the 1990s and both are now curiosities. People paid to have photographs digitised onto Photo CD specifically so they would last.

A Kodak Photo CD portable player, model PCD 880, photographed with its remote control, power adapter and SCART cable
Kodak sold players like this one so your photographs would always be viewable. The discs outlasted the hardware, the software and the company's interest. Photo: Hannes Grobe (CC BY 3.0).

Which format wins next is not something anyone can tell you, and anybody who claims to know is guessing. The useful move is not picking the winner. It is arranging things so that being wrong does not cost you anything.

What actually loses photographs

Format obsolescence is the risk people worry about. It is rarely what gets them.

What gets them is a single hard drive that fails, a phone that goes into the sea with the only copy of last year on it, a cloud account that lapses, a laptop that gets stolen, ransomware, or thirty thousand files named IMG_4471.jpg spread across four drives with no idea what is where. Storage media and human organisation fail far more often than formats do.

Two things matter more than your format choice:

Multiple copies, in multiple places. The standard formulation is three copies on two different kinds of storage with one of them offsite. It is a cliché because it works, and one external drive sitting next to your computer is not a backup, it is a second thing in the same room as the first thing.

Checksums, or at least redundancy. Silent corruption is real. A file can rot on a disk without anything reporting an error, and if your backup process faithfully copies the damaged version over the good one, you have lost the photo without ever seeing a warning. Generating SHA-256 checksums when you archive lets you detect that. Filesystems like ZFS and Btrfs do it automatically. If that is more than you want to take on, simply having several independent copies means a corrupted one is unlikely to be the only one.

The quiet killer: repeatedly re-encoding a lossy file. Every time you open a JPEG, make a small edit and save it again, it is re-compressed and loses a little more. Do that a dozen times over ten years of small tweaks and the degradation is visible. Edit from a master and export copies, rather than editing the same file over and over.

The rules that actually hold up

Everything above reduces to a handful of practical habits:

  1. Keep the unmodified original, whatever it is. The RAW, the camera JPEG, the scan. This is the one irreplaceable thing. Never edit it in place and never let a tool overwrite it.
  2. Do not re-encode lossy files without a reason. Converting a JPEG to another lossy format to save space costs quality permanently and gains you very little on the images you care most about.
  3. For anything you would be devastated to lose, hold a copy in a boring format. A baseline TIFF, or a maximum-quality JPEG, sitting alongside the original. This is your insurance against the original's format becoming awkward.
  4. Convert RAW to DNG if you want a documented container, but keep the originals too.
  5. Use modern formats freely for delivery and for everyday copies. AVIF, WebP and HEIC are excellent for the versions of an image that are supposed to be disposable. They are simply not where the last copy should live.
  6. Three copies, two media types, one offsite.
  7. Give files names and folders that mean something. Dates and events. Your future self will not remember, and no format decision helps with a file you cannot find.
  8. Open something old every year or two. A five-minute check that your oldest files still open is the only monitoring that matters, and it is how you find a problem while it is still fixable.

Doing something about it this week

You do not need to reorganise your whole library. Pick the folder you would be most upset to lose, and make sure that folder has an unmodified original of each image, a widely readable copy of anything that only exists in a proprietary format, and a second copy somewhere that is not your house.

For that widely readable copy, if the source is an ordinary image file, our image converter will produce a JPEG or PNG in the browser without your files leaving your machine, which is convenient when what you are doing is creating an archival duplicate rather than uploading originals to a service. The one thing it cannot do is read proprietary RAW, because browsers cannot decode those formats at all. Develop the RAW in desktop software first, export a TIFF or a high-quality JPEG, and let the converter handle any further copies you need.

Then set a reminder for two years from now to open the oldest file in that folder. That single habit will tell you more about the health of your archive than any amount of speculation about which format is going to win.