EXIF remover for photos and screenshots

Remove EXIF data from photos without posting them somewhere first

A photograph taken on a phone in the last decade almost certainly knows where it was taken. Latitude and longitude sit in the file at six decimal places, which narrows the spot to about the width of a dinner plate, along with the altitude and the second the shutter opened. That is the field people want gone before a picture goes up for sale, onto a forum, into a dating profile or out to a stranger who answered an advert. The customary way to delete it is to send the whole photograph, coordinate intact, to a website run by people you will never meet. The dangerous thing arrives at its destination and is then asked to leave.

Metadata removal cannot hide a face, address or serial number that is visible in the pixels. For that separate job, Redact image flattens deliberate black rectangles into a new PNG or JPEG.

If you only need to inspect a file, the image metadata viewer is read-only and shows its named EXIF fields without creating a replacement.

For lower-risk photographs where the surroundings should remain visible, local face blur proposes regions for review and then flattens pixelation or blur without sending the photo away.

Nothing here does that, and the arrangement is easy to verify rather than merely stated. Hand a picture in and the browser fetches one small program from this domain — the code that walks the blocks inside an image file — which is byte-for-byte the same for every visitor and is not fetched at all for someone who reads this page and leaves. After it arrives the network goes quiet. Open your developer tools and watch the request list as you work; better still, hand the file in so the program loads, pull the machine off the network, and do the job anyway. Checking a tool for yourself takes a minute and the method works on any site, including this one.

What these three formats can hold

JPG

Full name	Joint Photographic Experts Group
Extension	.jpg, .jpeg
Format type	Lossy raster image
MIME type	image/jpeg

PNG

Full name	Portable Network Graphics
Extension	.png
Format type	Lossless raster image, with alpha
MIME type	image/png

WebP

Full name	Web Picture format
Extension	.webp
Format type	Lossy or lossless raster image, with alpha
MIME type	image/webp

Nothing is re-encoded, and the receipt proves it

There are two ways to build a tool like this, and the difference shows up in your picture rather than in the marketing. The quick way is to decode the image and save it again: a handful of lines, and the common properties vanish because a fresh save writes none of them. It also pushes every pixel through JPEG quantization a second time, which softens edges, mottles smooth gradients and cannot be undone. Nobody tidying a photograph before publishing it has asked for a slightly worse photograph.

The other way treats the file as what it actually is — a chain of labeled blocks, some holding compressed picture data and some holding descriptions of you. Copy the first kind across untouched and omit the removable blocks the parser listed, and the pixels are not merely similar to the original, they are the identical bytes. When EXIF Orientation is 2–8, one tiny exception is required for the picture to keep looking the same: a newly built EXIF block carries that value and no other tag. That is the approach here, and because it is a factual claim it is checked rather than asserted: after every run the tool compares the compressed image data in the source with the compressed image data in the result and reports the outcome on the receipt. If you want a picture that is smaller as well as anonymous, that is a separate decision made deliberately with image compression rather than a side effect of cleaning.

Two ordinary blocks are deliberately left alone in a JPEG, for the same reason: they describe the picture, not the photographer. One records pixel density, so dropping it changes how large the image prints. The other tells decoders how the color channels are arranged, and files from certain professional workflows come back with inverted colors when it goes missing. Both are named in the list with a note saying they were kept, because a tool that quietly makes exceptions is a tool you have to trust instead of read.

C2PA credentials and AI labels are declarations, not detector verdicts

Provenance has more than one container. A JPEG can carry a C2PA credential in an APP11 JUMBF block; a PNG uses a caBX chunk. XMP can separately declare a digital source type such as trainedAlgorithmicMedia. This parser names those cases before the button and removes the whole containing block. Removing a credential breaks the chain a verifier would follow. It does not prove the image was fabricated, just as finding a declaration does not independently prove every claim inside it.

The format boundary matters. JPEG APP11 and PNG caBX are checked structurally. WebP EXIF and XMP chunks are removed, and explicit AI source terms inside XMP are identified, but this implementation does not claim general semantic C2PA/JUMBF coverage for WebP. The receipt's raw search knows several signatures and the values the parser extracted; it is a second mechanism, not a promise to recognize a container nobody has taught it.

The preview inside the picture, and the video nobody mentions

Two things live in image files that most cleaning tools never speak about, and both are worth knowing even if you use something else entirely.

The first is the embedded preview. A camera writes a small copy of the shot into the metadata so that a viewer can show a grid of images quickly, and editing software does not always regenerate it. Crop a face out of the corner of a photo and the preview may still show that corner. It is a documented cause of real disclosures, and it is invisible unless something goes looking. This tool reads the preview's dimensions and compares its proportions with the picture's — a mismatch is the fingerprint of exactly that situation, and it is reported in plain words before you run anything.

The second is everything after the end of the file. A JPEG has a defined final marker, and a phone shooting in motion or live mode appends a complete video after it. It is frequently the largest single thing in the file and no image viewer will ever show it to you. Here it is listed with its size, and it goes when you clean the file. If your pictures come off an iPhone they are probably HEIC rather than JPEG, in which case the honest answer is a different route: converting HEIC to JPG offers a metadata option, and leaving it unticked produces a JPEG with nothing in it.

What this EXIF remover will not do for you

It will not take a watermark off a picture. A visible mark is painted into the pixels; it is part of the image, and nothing described here touches pixels. What does disappear is an invisible ownership claim held in the properties — a copyright line, a credit, a rights statement — and that cuts both ways. If the photograph is yours, cleaning it removes your own attribution with the parsed block that contains it, so put the credit back in the caption where a human being will actually read it.

It will not write false information into a picture either. Tools exist that let you set a camera model, a date or a location after the fact, and there is precisely one decent use for that against several bad ones. We would rather not be the site that makes it easy to claim a photograph was taken somewhere it was not. And it will not tell you what a maker note contains: that block is the manufacturer's own undocumented format, different for every camera line, and the only honest report is its size — which is itself an argument for not publishing it. If your concern is a document rather than a photograph, the same job on the other side of the fence is stripping PDF metadata, and turning a picture into a plain web-ready file is PNG to JPG.