Remove Image Metadata

The readout arrives before the button does, on purpose. Deciding whether to strip a photo means knowing what is in it, and being told afterwards answers the question too late to be a decision.

1. Choose the photo you want cleaned. Drop a JPEG, PNG or WebP onto the panel above, or press Browse images. The file is read where it already sits, and all of it is read rather than the first few kilobytes, because a second picture can be stapled on after the first one ends.
2. Read what the file turns out to carry. The panel lists it under What this file carries: camera and capture data, location, an embedded preview thumbnail, editing history, captions and credits, comments, text chunks and anything appended after the picture, with a count beside a category found more than once. Only the categories are named — no value out of your file is put on screen.
3. Press Remove metadata. There is nothing to set. Every block that can go, goes; the colour profile stays so the colours still display the way they were meant to; and the compressed picture is copied across rather than re-encoded, which is why there is no quality control here.
4. Download clean image. The result says what was removed, what was kept and how much smaller the file is now. Press Download clean image to save it. The original on your device is left exactly as it was, so you can keep both.

A photograph is a container: the compressed picture sits in the middle of it and everything else is description written around the edges. Each kind of description has its own name, and the readout uses those names so you can see which ones your file actually has.

• Camera and capture data (EXIF). Make, model, lens, exposure, the capture date, and on many bodies a serial number that ties every photo you have ever published to the same camera.
• Location (GPS). Latitude and longitude, written by a phone by default. On a photo taken at home, indoors, this is a house number, and it survives being posted, mailed and re-saved.
• Embedded preview thumbnail. A second, smaller, genuinely visible copy of the picture, stored inside the EXIF block. It is generated when the photo is taken and has more than once shown what a later edit was covering up.
• Editing history, keywords and ratings (XMP). What software touched the file, sometimes every adjustment made to it, and any tags or star ratings from a photo library.
• Captions and credits (IPTC). Headline, caption, author, copyright and contact fields, written by newsrooms and by anything that has been through a picture desk.
• Extra data after the picture. Phones write more once the picture has ended, and every viewer ignores it: an Apple secondary image, which is a whole second photograph with its own camera data and its own coordinates, or the short video behind a motion photo. Those bytes are cut, along with the index that pointed at them.
• Comments, text chunks and the last-modified time. Free text a compressor or an exporter left behind, the name of the program that wrote the file, sometimes the folder path it came from, and the timestamp a PNG can carry inside itself.

The panel reports categories and counts and nothing else. No value out of your file is ever rendered, because a readout that helpfully printed a pair of coordinates on a shared screen would publish the exact thing you came here to remove.

The colour profile (ICC) is kept. A profile is not identity, it is colour: it tells a viewer what the numbers in the picture mean, and a photo edited in a wide-gamut space and then stripped of its profile is displayed by the wrong rulebook, so greens go flat and skin tones shift. Dropping it would change how the picture looks while telling you your privacy had improved. The readout names it and marks it as kept rather than pretending it was not there.

Two small JPEG headers stay for the same reason. One records the density the file claims, which is what a printer and a print shop read; the other tells a decoder how the colour channels in the scan are arranged, and removing it turns a normal photo into a colour bug rather than a private one. Need the print resolution set instead? That is a different job, on the same kind of header.

Every other tool here opens the picture, works on pixels and writes it out again. This one never opens it. A JPEG, a PNG and a WebP are all containers with a compressed payload inside, so only the container is rewritten: the descriptive blocks are left out and the compressed picture is copied straight across. What you download is the same picture, not a new rendering of it.

That matters because a JPEG loses a little every time it is saved. Open one, save it, open it again, save it again, and the detail wears down pass by pass, even at quality 100 — the encoder throws away the original coefficients and has to guess new ones. A tool that re-encoded to remove metadata would charge you a generation of quality for a privacy fix. Files here only get smaller, by exactly the weight of the blocks that came out. Need it smaller as well? Compress it afterwards, when losing some detail is the point rather than a side effect.

JPEG, PNG and WebP, up to 20 MB per file. HEIC is refused by name: rewriting its box tree is a different piece of work, and faking it by decoding and re-encoding would break the one promise this tool makes. Change the format first and the file that comes out is already clean, because a picture written from raw pixels has nothing to carry over. An animated PNG keeps its animation, which lives in chunks the picture needs rather than in the ones being removed.

This removes description, not content. Anything drawn into the picture itself stays: a visible watermark, a date burned into the corner by a camera, a face, a street sign, a screenshot of a document. Nothing here inspects pixels, so nothing here can find or change what is in them, and a stripped photo is not an anonymous one — it is a photo that no longer says which camera took it or where.

The one limit that comes from your own hardware is memory, and it is a light limit here. The file is walked as bytes and copied in pieces rather than unpacked into a full-size pixel buffer, so this is the cheapest job on the site — a 20 MB photo that a resize would have to think about costs little more than reading it.