Convert CR2 to PNG

PNG is genuinely lossless: what goes in comes out pixel for pixel. The catch on this pair is what goes in. A CR2 is not opened by demosaicing the sensor grid here — the converter finds the finished JPEG Canon wrote inside the file at capture and works from that. So the PNG is a lossless copy of a lossy original.

That is not a trick, it is arithmetic. Detail the camera’s JPEG encoder discarded is gone before PNG ever sees the picture, and no lossless container can put it back. What PNG does give you is a guarantee that nothing further is lost from this point on, which matters if the file is about to be edited and saved repeatedly, and matters not at all if it is going to be looked at once.

A 5D Mark IV frame is around 6,720 by 4,480 pixels — 30 million of them. At three bytes per pixel that is roughly 90 MB before any compression, and PNG’s compression works by finding repetition, which photographs contain very little of. Expect the output somewhere between 40 and 70 MB depending on the scene, with a dark, low-detail frame at the bottom of that range and a sunlit landscape full of foliage near the top.

The JPG from the same file would be a few megabytes. If that difference does not buy you something specific — an importer that refuses JPEG, an editing chain where every save would otherwise cost quality — it is a cost with no return, and the CR2 to JPG route is the better one.

There are real cases. Some competition and stock submission systems specify PNG and reject everything else. Some scientific and medical imaging pipelines require a format with no lossy step in it, regardless of what the source was. Some older design software has an importer that handles PNG cleanly and JPEG badly. Some print workflows have a house rule and no interest in your reasoning about it.

In all of those the answer is the same: produce the PNG, do not argue with the form, and keep the CR2. What you should not do is convert to PNG because it sounds like the higher-quality option — on this pair it is the same picture in a larger container.

Two things go. The first is everything the sensor recorded beyond the camera’s own rendering: the 14-bit tonal range, the highlight headroom, the ability to reset white balance after the fact. That is in the CR2 and it is not in the extracted preview, so it is not in the PNG either.

The second is the metadata. Making a PNG requires decoding the preview and encoding it again, and the EXIF block does not survive that. The lens, the exposure settings, the capture time and any GPS tag are absent from the output. If you want that information kept, either keep the CR2 alongside or use the JPG route, which passes the preview through without re-encoding it — and there, turn off Remove metadata, because it is on by default and strips the same block.

PNG’s best-known advantage over JPEG is the alpha channel, and it is irrelevant here. Every pixel in a raw capture was recorded by a photosite and is fully opaque; there is nothing transparent in the file for PNG to carry.

If you need a cut-out — a product on white, a subject on a transparent background — that is a masking job in an image editor after the conversion, not something the conversion can do. The PNG gives you a container capable of holding transparency; it does not create any.

PNG can hold 16 bits per channel, which sounds like a natural home for a 14-bit Canon capture. It is not, because the 14-bit data never reaches the encoder. The embedded preview is an ordinary 8-bit JPEG, so the PNG written from it is 8 bits per channel with no more tonal range than any web image.

A 16-bit PNG that actually carried the sensor’s range would have to come from a raw developer that demosaics the CR2 — Digital Photo Professional, Lightroom, Camera Raw or darktable can export exactly that. That is a different piece of software doing a different job, and it is worth naming rather than implying this page does it.

Large PNGs run into limits that JPGs never touch. Email attachments commonly cap at 20 or 25 MB. Content management systems frequently reject uploads above 8 or 10 MB by default. Messaging apps re-compress anything large and hand the recipient a worse picture than the JPG would have been.

The free tier here accepts input files up to 100 MB, which every Canon CR2 fits inside comfortably — the constraint is on the output and on whatever you are sending it to. It is worth checking the destination’s limit before converting a batch, because discovering it after producing forty 60 MB files is an annoying way to learn.

Batches work: drop a selection and each file is scanned, extracted and re-encoded in turn, with the finished set available as a ZIP. Re-encoding is real work, though, and PNG compression on a 30-megapixel image is not instant, so a large batch of PNGs takes noticeably longer than the same batch as JPGs.

It also runs entirely on your machine, which is the reason there is no per-file charge and no queue, and the reason a shoot that has not been published yet does not have to be handed to a website first.

If PNG was chosen because you intend to work on the picture, the honest recommendation is to skip this page. Open the CR2 in a raw developer, make the exposure and white balance decisions against the sensor data where they are nearly free, and export a TIFF or a 16-bit PNG from there.

This converter exists for the case where that is not available or not worth it — a machine without the software, a file needed in the next ten minutes, a destination that only wants a picture. Within that case it does the job well. Outside it, it produces a large file that a better tool would have produced better.