RAW formats compared: what actually differs between ARW, CR3, NEF, RAF and the rest

Every camera maker has its own raw format, but underneath they store the same kind of data. The real differences are in packaging, compression options, software support and how safe they are to archive.

Ask why a Sony file ends in .ARW and a Nikon file in .NEF and the honest answer is mostly history. Each manufacturer designed its own container in the early days of digital photography, and each kept refining it as sensors changed. The pixels inside are fundamentally alike: values from a sensor covered by a color filter array, plus metadata describing the camera, the lens and the shooting settings.

That does not make the formats interchangeable, though. They differ in how the data is packaged, which compression choices you get in the camera menu, how quickly new models are supported by third-party software, and how confident you can be that a file will still open in twenty years. This guide walks through those differences without the marketing.

The big comparison table

"Typical" matters in that table: options vary by model and firmware, so your camera's menu is the final word.

What all raw formats share

Strip away the containers and every format above stores the same basic ingredients:

• Sensor values, usually 12 or 14 bits per photosite, before demosaicing, white balance or a tone curve have been applied.
• Metadata: camera and lens model, exposure, the as-shot white balance, and maker-specific notes such as picture style or film simulation choices.
• One or more embedded JPEG previews, which is what your camera screen and many file browsers show you.

The embedded preview is the source of a common misunderstanding. The camera's look, whether Canon Picture Style, Nikon Picture Control, Fujifilm film simulation or a LUMIX Photo Style, lives in that preview and in the maker's notes, not in the sensor data. A third-party raw developer shows a neutral rendering, which is why raw often looks flatter than the camera's JPEG. See why raw looks flat for the full explanation.

Container differences that actually matter

TIFF-based formats

ARW, CR2, NEF, PEF and DNG are all built on TIFF, a structure of tagged data blocks that dates back to the 1980s. ORF and RW2 are TIFF-like variants with their own quirks. A shared base made it relatively easy for open-source developers to read them; the hard part was always the undocumented details inside the maker notes and the compression.

Canon CR3

CR3 broke with tradition by using the ISO base media file format, the same family as MP4 and HEIF. That made it a natural fit for cameras that also record video, but at launch it meant software had to add a whole new decoder rather than tweak an existing one. Support took a while to spread, which is a good example of why brand-new formats carry a short-term risk.

Fujifilm RAF

RAF uses Fujifilm's own container. The bigger difference is the X-Trans sensor used in most X series cameras: its 6x6 color pattern needs a different demosaicing approach from the 2x2 Bayer pattern everyone else uses. Results from generic developers can differ more from the camera's JPEG than with other brands, especially in fine foliage detail. GFX medium format cameras use Bayer sensors.

DNG

DNG's selling point is the public specification. Anyone can write a reader without reverse engineering, and the format can describe both untouched mosaic data and already demosaiced "linear" data. That flexibility is why phones use it, and also why a phone DNG and a Leica DNG can behave very differently in the same software.

Compression: lossless, lossy and "efficient"

Raw compression comes in two kinds, and makers name them inconsistently.

• Lossless compression shrinks files without changing any value, much like a ZIP file. There is no image-quality reason to avoid it. Canon's standard RAW, Nikon's Lossless compressed, Sony's Lossless compressed and Fujifilm's Lossless compressed all fall here.
• Lossy compression throws some information away in exchange for much smaller files. Canon C-RAW, Nikon's Compressed and High Efficiency modes, Sony's Compressed, and Fujifilm's Compressed are lossy to varying degrees. In normal scenes the difference is hard to see; it tends to show up, if at all, in extreme shadow lifting or very high-contrast edges.

A useful rule of thumb for storage planning: an uncompressed 14-bit raw needs about 1.75 bytes per pixel, so a 24-megapixel camera writes roughly 42 MB per frame before the preview and metadata. Lossless compression typically brings that down noticeably, and lossy modes further still; exact savings depend on the scene and the camera.

Newer "high efficiency" codecs, such as Nikon's High Efficiency NEF on recent Z bodies, are the ones most likely to be unsupported in older or less common software for a while after release. If you depend on a particular tool, check its support before switching your camera to a new compression mode.

Software support in practice

Three groups of software read raw files:

1. The maker's own converter: Sony Imaging Edge, Canon Digital Photo Professional, Nikon NX Studio, Fujifilm X RAW Studio, OM Workspace and SILKYPIX for Panasonic. They are free, support new cameras on day one, and reproduce in-camera looks most faithfully. They usually only read their own brand.
2. Commercial editors such as Lightroom and Capture One, which add new cameras in regular updates, usually within weeks or a few months of release.
3. Open-source decoders, chiefly LibRaw (the successor to Dave Coffin's dcraw) and rawspeed, which (alongside each project's own code) power darktable, RawTherapee and many smaller tools, RAWSnap included. They cover an enormous range of cameras, past and present, but newly released models can take longer to be added.

So the practical difference between formats is less "which is better" than "how long after buying a new camera can my favorite tool open its files." For a common format on a popular camera, the wait is short. For a new container or codec, it can be longer. The raw editors comparison goes deeper into the tools themselves.

Longevity: will these files open in 2045?

Proprietary formats are undocumented, which sounds alarming, but in practice popular formats have proven durable. Files from early-2000s DSLRs still open in current software because LibRaw and similar libraries keep support for old cameras, and those libraries are open source, so the knowledge is not locked inside one company.

The risk is real but concentrated in niche cases. Formats used by few cameras, or ones that are unusually complex, get less attention. Sigma's Foveon X3F files are the classic example: some open-source editors have dropped support for them. Very new codecs are a smaller, temporary risk.

Sensible archiving habits cover both cases:

• Keep your original raw files untouched, whatever else you do.
• Consider DNG copies for formats from niche or discontinued systems, since an open specification is the best long-term insurance.
• Export a high-quality JPEG of your keepers. JPEG will almost certainly be readable for decades, even if it lacks raw latitude.
• Keep at least two copies on different media, one off-site.

The archiving guide turns this into a concrete folder and backup plan.

Should you convert everything to DNG?

Not necessarily. Converting gives you an open, widely readable file with the same image data. It also takes time, the maker's own software usually will not open the converted file, and some maker-specific metadata may not carry over. Many photographers keep originals and only make DNG copies for older or less common cameras. If your camera offers DNG natively, as Pentax bodies do, choosing it at capture avoids the extra step entirely.

Which format is "best"?

The format your camera writes is the one you should keep. Image quality is set by the sensor, lens and exposure, not by the container. The meaningful choices are the compression mode (lossless is the safe default if card space allows) and, on Pentax, PEF versus DNG. Whatever you shoot, a browser-based converter like RAWSnap can turn the files into JPEGs locally when you need something shareable, and a full editor is there when a frame deserves more work. If you are still deciding whether raw is worth it at all, start with RAW vs JPEG.

FAQ

Is one raw format higher quality than another?

No. Quality depends on the sensor and how you expose. The container and lossless compression do not change the image. Lossy compression modes can lose a little information in extreme edits.

Why does Canon have both CR2 and CR3?

CR2 was used by Canon DSLRs for many years. From 2018 Canon moved to CR3, which uses a different container and added options such as the smaller C-RAW mode.

Can I rename an ARW to DNG to make it open?

No. The extension only labels the container; renaming does not change the data. Use Adobe DNG Converter to create a real DNG.

Is DNG safer for archiving than proprietary formats?

Its public specification makes it the safest bet in principle. In practice, popular proprietary formats have also remained readable for decades. Keeping originals plus a JPEG of your best shots covers most risks.

Why won't my brand-new camera's files open?

Third-party decoders need to add each new model. Use the manufacturer's free converter until support arrives.