The JPEG hiding inside your RAW file

Your camera already rendered every frame you shot, once, at the moment of capture, and put the result inside the RAW file. That is what you see on the back screen, and it is what any viewer shows you when a RAW opens "instantly". Knowing how big it is on your brand, and how to pull it out, saves a develop you often do not need.

What is in there and why

A RAW cannot be shown without a decoder, and a camera's screen needs to show it instantly, so every camera renders a JPEG at capture and stores it inside the RAW container alongside a small thumbnail. The render is the camera's own: Picture Style, white balance, sharpening, sometimes lens corrections. In TIFF based formats (CR2, NEF, ARW, DNG, PEF, ORF, RW2) it sits in an image directory the file points to. In Canon's CR3 it sits in its own box in the ISO container. In Fujifilm's RAF it sits right at the front of the file, behind a short header. Either way it is a complete, standalone JPEG, and pulling it out is a copy, not a conversion.

Sizes by brand

This is the part that decides whether the embedded JPEG is deliverable. The numbers below are typical for recent bodies; the tool in the next section tells you exactly what your file carries.

The Sony number surprises people most. A 61 megapixel a7R V carries a preview smaller than a phone screen. It is a fine preview; it is not a print.

How to extract it in seconds

Drop the file on Kepla's RAW to JPG converter. Every file shows its embedded JPEG in under a second, and the viewer tells you its exact pixel size next to the file name. Set the mode to Camera preview and download: the JPG you get is the camera's render, copied out of the RAW with a clean EXIF block written in, GPS left out. Forty files at a time, one ZIP, nothing uploaded. If the size on the label is smaller than you need, switch to Full develop and the same page renders the sensor data at full resolution instead.

The command line crowd has done this for years with exiftool (-PreviewImage or -JpgFromRaw, depending on the brand) and dcraw's -e flag. Same result, fewer sliders.

When it is enough, and when it is not

Enough: a sneak peek for the couple before the reception ends, a card check, a proof for a client who wants to see a frame, a social post, anything where the camera's look is the look you want and the size on the label is at least the size you need. It is instant and it is exactly what the camera showed you.

Not enough: a print, a delivery from a Sony, Olympus or Panasonic body, a frame that needs an exposure nudge or a white balance fix, or any frame where the camera's Picture Style is not the look you want. That is a develop, from the sensor data, and it costs a few seconds per file. How to convert RAW to JPG covers every route.

The culling angle

The embedded JPEG is also why a culling pass can be fast. Whether eyes are closed, whether a burst is duplicates, whether the frame is empty, whether it is badly out of focus: those answers are the same at preview size as at full resolution, and reading a preview is up to forty times cheaper than decoding the RAW. That is how Kepla clears the obvious misses from 4,812 frames in 4 minutes 22 seconds on a MacBook Pro: it reads the preview where the answer is clear and opens the RAW where it is not, 214 frames on that wedding. Why culling should read the embedded JPEG has the full argument, and Founding Access is how you get the app.

FAQ