How to Convert CR3 to JPG

Both routes end in a JPG, but they are not the same operation and they do not produce the same file.

Extracting the camera's rendering takes the full-size JPEG that your Canon wrote inside the raw file when you pressed the shutter, and re-encodes it at the size and quality you ask for. It carries the picture style, white balance and lens corrections that were active at capture. It is close to instant, needs no installation, and gives you the frame you saw on the rear screen.

Developing the raw data reconstructs the image from the sensor measurements, which is what Digital Photo Professional, Lightroom, Capture One, darktable and RawTherapee do. Slower, needs software, and buys you the thing the format exists for: highlight recovery, a white balance you can still change, and shadow detail that has not already been decided.

The test is simple. If you want to deliver these photographs — proofs, a website, a gallery, an email to the person who asked for them — extraction is the right tool and the rest of this guide is for you. If you want to edit them, and particularly if any frame needs rescuing, open a raw editor instead and treat the JPG as an export from that.

Nothing is installed and no file is uploaded — the parsing and the JPEG encoding both happen on your own machine, which is worth knowing if the frames are a client's or the shoot is not public yet. You can verify that claim the blunt way: load the page, disconnect from the network, and convert anyway.

These two settings decide the file size you end up with, and they do it in very different ways.

Quality runs from 10 to 100 and controls how hard the JPEG encoder compresses. The useful range is narrower than the slider suggests. Around 90 is a sensible default for anything you might print or hand over: the compression is effectively invisible and the file is a fraction of the raw. Around 80 is right for the web, where nobody is inspecting at 100 percent. Below about 60, flat areas like sky start showing blocky patches, and no amount of later processing will get them back.

Output size offers the original dimensions, 75 percent, 50 percent, or custom pixel dimensions with the aspect ratio locked by default. Reducing dimensions is a far blunter instrument than reducing quality, and much more effective — halving the long edge quarters the pixel count. For proofs and web galleries, 50 percent at quality 85 usually produces a file a tenth the size of a full-resolution export that nobody can tell apart at normal viewing distance.

A worked example, for a 45-megapixel frame from an R5:

If you need to hit a specific ceiling — under 500 KB for an upload form, say — convert generously here and then compress to a target size, which steps the quality down until the file fits rather than making you guess.

By default the camera information travels with the photograph. The converter reads the Exif blocks out of the raw file and rebuilds them into the JPG: make and model, shutter speed, aperture, ISO, focal length, lens, capture date, and GPS coordinates if your camera recorded any. Each finished frame lists what it found, so you can see before downloading.

Two things do not survive. Canon's proprietary maker notes are dropped, because their internal offsets only make sense inside the original container. And the output is stamped as having been produced by this converter, which is ordinary practice for anything that writes a JPEG.

There is also a switch to strip metadata entirely. Use it when the destination is public. A converted frame otherwise announces the camera body, the lens, the exact timestamp and — for anything shot with a phone-linked or GPS-equipped body — the coordinates you were standing on. That is fine for a client delivery and not fine for a photograph of your house going onto a forum. If you have already converted with the tags kept, stripping them afterwards is a second pass rather than a re-conversion.

The orientation switch is worth understanding too. Portrait frames are stored as landscape data plus a rotation tag, and not every viewer honours that tag. Leaving auto-orientation on physically rotates the pixels and rewrites the tag to normal, so the JPG appears the right way up everywhere rather than only in Exif-aware software. Turn it off only if you specifically want the raw orientation preserved.

Select every file at once rather than working through them one at a time — the settings are applied to the batch, so you configure once.

Files convert one after another rather than in parallel, which sounds slower and is not: the bottleneck is JPEG encoding, and running several at once on the same machine mostly produces memory pressure. Each row reports its own progress and its own outcome, so a single problem file does not take the batch down with it. You can stop the queue partway through and keep everything finished up to that point.

When it is done, download frames individually or take the lot as a single ZIP. For a card of a few hundred frames, work in groups of thirty or forty; the limit is the memory your browser tab can hold at once, not the tool.

One caution on very high-resolution output. Browsers cap how large a canvas they will render, and a full-size frame from a high-megapixel body can approach it. If a conversion fails at original size, drop to 75 or 50 percent and it will go through — the tool also steps the size down itself before giving up.

Converting when you meant to edit. The single most common one. Extraction gives you the camera's rendering; if the shot needs its highlights pulled back, that has to happen in a raw editor before anything becomes a JPG. Converting first and editing the JPG afterwards means editing eight bits of data that has already thrown away the latitude you needed.

Deleting the raw files afterwards. The JPG is the decision made permanent. Card space is cheap, and the measurements in a CR3 cannot be reconstructed from its JPG.

Converting at full size for the web. An eight-megabyte hero image is a slow page and, on a metered connection, a rude one. Convert, then resize and compress.

Assuming a failed file means a broken tool. A .CRM video file, a truncated copy from a failing card, or a file that never finished writing will all refuse. The specific failures and what each one means are worth two minutes before you retry.

Leaving GPS tags on a public upload. Easy to forget, occasionally consequential.