What a RAW file is — and the JPG already inside it
A RAW file is not a photo. It is the unprocessed reading from your camera's sensor, plus a pile of metadata about how the camera was set when the shutter fired. Nothing in it is a picture yet. Turning it into one — demosaicing the Bayer pattern, applying the maker's colour science, sharpening, correcting the lens — is called developing, and it is the reason a 24-megapixel frame arrives as a 30 MB file instead of a 5 MB one. You keep all that data so you can change your mind later about white balance, shadows and contrast.
Here is the part most people never hear: every camera also writes a finished, ordinary JPEG inside that same RAW file. It has to. The rear screen needs something to draw, and the screen cannot develop a RAW any faster than a laptop can. So the camera develops the shot once, right at capture, using the picture style and white balance you had set, and tucks the result inside the file as a preview.
How big that hidden JPEG is
On most bodies made since roughly 2012, it is the full sensor resolution. A 24-megapixel camera writes a 6,000 × 4,000 JPEG. Some older and entry-level cameras write a screen-sized one instead — often 1,600 × 1,200 — which is fine to look at and too small to print.
What that means for converting
Pulling that JPEG out is not the same as developing the RAW, and anyone telling you otherwise is selling something. It is faster, it is lossless, and for most of the reasons people want to convert raw to jpg — proofs, a quick look, sending shots tonight — it is the right answer. That is what the RAW to JPG tool on this page does: find the JPEG, copy it out whole, hand it to you.
How RAW to JPG conversion works in your browser
No browser ships a decoder for CR2, NEF or ARW. That is not a gap someone will patch next year; it is camera-maker colour science, and it stays proprietary. So any converter that runs in a tab is doing one of exactly two things: uploading your photos to a server and developing them there, or reading the JPEG the camera already wrote. This tool does the second one, in the open.
It reads the first megabyte of each file, walks the format's index to find where the embedded JPEGs live, picks the largest one, copies those exact bytes into a new file, and hands you a download link. No pixel is decoded. No pixel is re-compressed. Nothing goes to a server — which is why it keeps working with the Wi-Fi off, and why a 2 GB batch has no upload to wait through.
There is no formula here, and that is the whole point. Converting usually means decode, then re-encode, and re-encoding a JPEG always costs a little quality. This copies a run of bytes out of one file and into another. The output is bit-identical to what your camera made.
A worked example. Drop in IMG_0431.CR2, a 25,804,800-byte file from a Canon body. The row reads Canon CR2 · 24.6 MB · waiting, then a moment later: JPG 3.1 MB · 5,472 × 3,648 px · 88% smaller than the RAW. The download button says IMG_0431.jpg. That is a 20 MP JPEG, extracted in about a second, from a file your browser cannot actually decode.
Which RAW formats work, and the one that doesn't
Most RAW formats are TIFF files wearing a different extension, so one reader handles the bulk of them. A few makers do their own thing and need their own path. Either way you pick nothing: cr2 to jpg, nef to jpg, arw to jpg and dng to jpg all run down the same road, and the RAW to JPG tool works out which one from the file's first bytes.
| Format | Extensions | Where the JPEG hides | What you usually get |
|---|---|---|---|
| Canon CR2 | .cr2, .crw | A JPEG-compressed single strip in the TIFF index | Full sensor resolution |
| Canon CR3 | .cr3 | A different container entirely — found by scanning for JPEG markers | Full sensor resolution |
| Nikon NEF | .nef, .nrw | The standard EXIF offset-and-length pair | Full sensor resolution |
| Sony ARW | .arw, .sr2, .srf | The standard EXIF offset-and-length pair | Full sensor resolution |
| Adobe DNG | .dng | A preview sub-index inside the TIFF structure | Depends on what wrote the DNG |
| Fujifilm RAF | .raf | Fuji's own header, two big-endian numbers at byte 84 | Full sensor resolution |
| Panasonic RW2 | .rw2, .raw | A dedicated tag, 0x002E, whose length is the JPEG's length | Full sensor resolution |
| Olympus ORF | .orf | TIFF layout with Olympus's own magic number | Often screen-sized on older bodies |
| Everything else | .pef, .srw, .erf, .mrw, .rwl, .dcr, .kdc, .3fr, .fff, .iiq, .cap, .mef, .mos, .ari | TIFF index, same reader | Varies by camera |
| Sigma X3F | .x3f | Nowhere — Foveon files carry no JPEG preview | Refused, with the reason on the row |
Sigma X3F is the honest failure. Rather than churn and produce nothing, the row says so plainly: Sigma X3F files don't carry a JPEG preview to pull out. Sigma Photo Pro can export one. That is the only common format with no path through this tool.
Extensions are a hint, not evidence. The tool reads the first bytes and trusts those, so a .dng that is secretly something else, or a CR2 someone renamed, is judged on what it actually contains.
What actually comes out: sizes from a real batch
Eight files from eight different bodies, dropped in as one batch. Every figure below is what the RAW to JPG tool reports on the row, not an estimate.
| File | Format | RAW size | JPG out | Dimensions | Smaller by |
|---|---|---|---|---|---|
| IMG_0431.CR2 | Canon CR2 | 24.6 MB | 3.1 MB | 5,472 × 3,648 px (20 MP) | 88% |
| IMG_0431.CR3 | Canon CR3 | 28 MB | 3.5 MB | 6,000 × 4,000 px (24 MP) | 88% |
| DSC_7781.NEF | Nikon NEF | 30 MB | 2 MB | 6,000 × 4,000 px (24 MP) | 93% |
| DSC01204.ARW | Sony ARW | 48 MB | 5 MB | 7,952 × 5,304 px (42.2 MP) | 90% |
| DSCF4417.RAF | Fujifilm RAF | 52 MB | 4 MB | 7,728 × 5,152 px (39.8 MP) | 92% |
| IMG_2210.DNG | Adobe DNG | 25 MB | 2.5 MB | 4,032 × 3,024 px (12.2 MP) | 90% |
| P1000876.RW2 | Panasonic RW2 | 23 MB | 3 MB | 5,184 × 3,888 px (20.2 MP) | 87% |
| P1010324.ORF | Olympus ORF | 17 MB | 240 KB | 1,600 × 1,200 px (1.9 MP) | 99% |
247.6 MB of RAW went in. 23.3 MB of JPG came out. The pattern holds across makers: expect the JPG to be roughly a tenth of the RAW, give or take, at the same pixel dimensions.
The Olympus row is the one to read carefully. 99% smaller looks like a win until you see 1,600 × 1,200 px next to it — that camera stored a screen preview, not a full frame. The tool flags it before you download: Your camera stored a 1,600 × 1,200 preview, not a full-size one. Good for a quick look, not for printing. A suspiciously small JPG is the tell, and the dimensions confirm it.
When to pull the JPG, and when to develop the RAW
Pull the JPG when you need the picture and you need it now. A wedding second-shooter hands you a card and the couple wants a teaser tonight. A client is waiting on proofs. You are on a borrowed laptop that will never have photo software installed on it. Someone emailed you forty CR2 files and you just want to see what is in them. In every one of those cases, the camera already made the image you want — going back to the sensor data buys you nothing.
Develop the RAW when you intend to change it. Recovering two stops out of a blown sky, fixing white balance under mixed light, pulling detail out of shadows, applying a profile that isn't the one the camera had loaded — that is sensor-data work, and no extraction can fake it. Use the software that came with your camera, or whatever you edit in. It will beat this, and it is supposed to.
The gap worth naming is the third case: the shape most tools for this job take. A 700 MB install tied to one machine. Or a web page that quietly uploads your client's photos to somebody's server, stamps a watermark across the result, and offers to remove it for a monthly fee that renews whether you shot anything that month or not. Or a free tier that caps you at three files and calls the fourth an upgrade. None of that is required to copy a JPEG out of a file that already contains one. So the RAW to JPG tool doesn't do any of it: no account, no upload, no watermark, no cap beyond what your own machine can carry.
Edge cases and gotchas
A few behaviours are worth knowing before you point a full card at it. Batch raw to jpg work is where the small decisions show up.
- Duplicate names get suffixed, never overwritten. Add IMG_0431.CR2 and IMG_0431.CR3 to one batch and you get IMG_0431.jpg and IMG_0431-2.jpg. Matching is case-insensitive, so img_0431.cr2 counts as the same name.
- Only image extensions are stripped. shoot.v2.NEF becomes shoot.v2.jpg — the .v2 is part of your name, not a file type, so it stays.
- The limits are 50 files, 300 MB per file, 2 GB per batch. Go over and the row tells you which: Too large: 350 MB. Max is 300 MB per file. Clear the finished files to make room and keep going.
- A truncated card copy fails cleanly. If a file copied across incompletely, its internal offsets point at nothing. The row says so and suggests recopying from the card, rather than handing you a broken download.
- Files that are already JPGs are skipped, not re-saved. This is already a JPG. There's nothing to pull out.
- Empty files are caught. A 0 B file is a copy that didn't finish, and it gets flagged as one.
Nothing in the batch stops the rest of the batch. Bad files are marked with a reason and skipped; the good ones convert around them.
Related image tools
If the file you are holding is not a camera RAW, one of these probably fits better. Phone photos from an iPhone are HEIC, not RAW — the HEIC to JPG converter handles those. Images saved out of a browser are often WebP; the WebP to JPG converter takes those. Email attachments that arrive as .jfif are JPEGs with a confusing extension, which the JFIF to JPG converter sorts out. For anything else, the image format converter moves between the common formats.
Once you have your JPGs, the image compressor will shrink them for email or the web, and the image resizer will bring a 42-megapixel extraction down to something a website will accept. All of them run in the browser, the same way this one does.
Frequently asked questions
Does the JPG keep my EXIF and shooting data?
Whatever your camera wrote into that embedded JPEG travels with it, because the bytes are copied out whole. The tool adds nothing and strips nothing. What it cannot do is move metadata that lives elsewhere in the RAW into the JPG — if a field was stored outside the preview, it stays there. In practice most cameras include the usual shooting data in the preview, so aperture, shutter and ISO normally come through.
Why did two photos from the same shoot give me very different JPG sizes?
JPEG compression is content-dependent. A frame full of fine detail — foliage, crowds, texture — needs more bytes than a portrait against a plain wall, even at identical settings and dimensions. A 3.1 MB JPG and a 2 MB JPG from the same camera are both normal. The number to check is the pixel dimensions on the row, not the file size. If the dimensions match your sensor, you have the full frame.
Do my original RAW files get changed?
No. The tool only reads. It opens each file, copies a range of bytes out, and never writes back. Your RAWs on disk or on the card are untouched, and the JPGs land wherever your browser puts downloads. There is nothing to undo.
What happens if two files have the same name?
The second one gets -2, the third -3, and so on. This matters more than it sounds: two cameras shooting the same day produce colliding names constantly, and a folder of DSC_7781 files from three bodies would otherwise silently overwrite each other on download. Nothing in a batch can overwrite anything else in that batch.
Can I do this on an iPad or a Chromebook?
Yes. It needs a modern browser and nothing else — no install, no extension, no companion app. That is most of the point: the machines least likely to have photo software on them are exactly the ones you end up holding when someone hands you a card. Large batches are limited by the device's memory more than anything, so a phone will handle a few dozen files rather than fifty large ones.
Is there a watermark, a file cap, or a paid tier?
None of the three. The JPG you download is the camera's own bytes, unmarked. The only caps are the technical ones — 50 files, 300 MB each, 2 GB per batch — and they exist because a browser tab has finite memory, not because there is a bigger version to sell you. Microapp is paid for by memberships and ads across the site as a whole, and gives 10% of every dollar it earns to charity, off the top and audited quarterly. This page is simply the tool.