Convert CR3 to PNG

A CR3 is a container of sensor readings with Canon-specific structure, introduced in 2018 and built on the ISO base media file format rather than on TIFF. Placing a picture in a document requires the software to decode it, and Word, PowerPoint, Slides, Confluence, Notion and Figma have no raw decoder in them. They do not fail gracefully with raw files; they usually refuse to see them at all.

PNG is the opposite case: a W3C Recommendation from 1996, an ISO standard since 2004, that every one of those tools has read since it existed. Converting is not a preference here, it is the only way the picture gets into the document.

The converter does not develop the sensor data. Every Canon writes a complete JPEG inside the CR3 at the moment of capture, at full resolution on almost every body, and that is what is found and re-encoded as PNG. It carries the picture style the camera had set and the white balance the camera chose.

For a photograph going into a report or a deck this is not a compromise — it is the same picture the photographer looked at on the camera screen and said yes to. It only becomes a problem if the frame needs rescuing, and a document is rarely the place where that conversation is happening.

This is the failure people actually hit. Forty-five million pixels at three bytes each is about 135 MB before compression, and PNG compresses photographs poorly because it works by finding repetition and photographs contain almost none. A single full-size PNG can land above 70 MB.

Put six of those in a presentation and the file will not attach to an email, will crawl when scrolled, and will make the shared drive unhappy. PowerPoint and Word will compress images on save if you let them, which produces a JPEG inside the document and quietly undoes the reason you chose PNG in the first place.

The fix is upstream of the format choice. A full-page image in a document is printed or displayed at perhaps 2,000 pixels across; a half-column figure at 900. Setting a maximum width when you convert removes the pixels nobody will ever see and takes the file from tens of megabytes to a few.

It has one side effect worth knowing: resizing forces the picture through a decode and re-encode, which is already the case for PNG, so on this pair you lose nothing by using it. Choose the width the document actually needs and the size problem disappears.

PNG is often chosen because somebody remembers it as "the one with transparent backgrounds". That is true of the format and untrue of this file. Every photosite on the sensor recorded a value, so every pixel is opaque, and the alpha channel PNG provides arrives empty.

If the deck needs a product on a transparent background, that is a masking step in an editor — or a photograph taken against a background you can key out — and it happens after this conversion rather than during it.

PNG can store 16 bits per channel, and a Canon sensor records 14. It is tempting to assume the two meet. They do not, because the 14-bit data is never decoded here — the embedded preview is an ordinary 8-bit JPEG, so the PNG written from it is 8-bit as well.

This matters if someone downstream intends to grade or heavily adjust the image, because 8-bit data bands when it is pushed hard. If that is the plan, the CR3 should go to Digital Photo Professional or Lightroom and come out as a 16-bit TIFF, and this page is not the route.

The EXIF block does not survive the re-encode. Camera model, lens, exposure settings, capture time and any GPS coordinate are absent from the PNG. For an internal document that is neutral; for anything published it is a small piece of accidental data protection, since office cameras and phones both tag location more often than their owners realise.

It also means attribution has to be handled in the document rather than in the file. If the photographer needs crediting, the caption is the place, because nothing in the PNG will carry their name.

Batches are the normal case here: a shoot arrives as forty frames and six of them are wanted. Drop the selection, let each one extract and encode, and take the result as a ZIP. The free tier accepts files up to 100 MB, which every CR3 fits under — a full-raw R5 frame is around 45 MB and a C-RAW frame considerably less.

PNG encoding is real work, unlike the JPG route where the preview is copied out untouched, so a large batch of full-size PNGs takes a while. Setting a maximum width speeds that up as much as it shrinks the files.

The conversion runs in the browser, so the CR3 files never leave the machine. That is worth something in an office context specifically: photographs of staff, of a customer site, of an unannounced product, or of anything covered by a client agreement should not be uploaded to a converter chosen from a search result.

You can verify it rather than believe it. Open the developer tools, watch the network tab while a file converts, and note that nothing carrying the image is sent.