Remove Metadata from PDFs and Photos — Free & Private

Two things, and it says how many of each it found before you press anything. The first is the /Info dictionary — the document properties a reader shows under File then Properties: title, author, subject, keywords, the creating application, the producing library, and the creation and modification dates. Every key is deleted outright rather than set to an empty string, because an emptied title is still a title. The second is XMP: the same kind of information as an XML packet, which can sit on the document, on an individual page, and on photographs embedded in the file. Clearing the properties alone is the common half-measure — an XMP packet routinely repeats the title and the author, so a document cleaned that way still names you. Both go, and the objects they lived in are dropped from the file rather than left in it unreferenced.

Because a browser canvas cannot write those containers. Cleaning an image works by decoding the picture and encoding it again — that is what leaves the metadata behind — so the output can only be a format the browser's encoder supports. PNG stays PNG and WebP stays WebP; JPEG, HEIC, HEIF and TIFF are all written out as JPEG. The page states the container it plans to write before you run it, then reads back what the encoder actually produced, so if a browser quietly substitutes a format — WebKit writes PNG when it is asked for WebP — you are told, instead of ending up with a .webp file full of PNG bytes.

Yes, and that is the part worth knowing. The download is a new picture built from your pixels, not your original bytes with the tags cut out. JPEG and WebP are written at the quality you choose, 92% by default, which gives up a little detail; the embedded colour profile is not carried over, because the canvas works in sRGB; anything deeper than 8 bits per channel comes back at 8; transparency survives only in PNG and WebP, since a JPEG has no alpha channel and transparent areas are filled white; an animated PNG or WebP keeps only its first frame; and a multi-page TIFF keeps only its first page. Keep the original if the pixels matter. A PDF is the opposite case — nothing on its pages is re-encoded, so the document looks and prints exactly as it did.

Some things, and they are named here rather than glossed over. In a PDF the pages themselves are untouched, so a name typed into a page, an annotation's author or a form field's value is still there — that is content, not metadata. Any /ID strings in the trailer, which identify the file rather than describe it, are left in place as well. And if your document is a PDF/A archival file, PDF/A identifies itself through an XMP packet, so removing the metadata removes that identification too. For an image, the listing covers what this page's reader can parse — EXIF, GPS, IPTC, XMP and the colour profile — and not every block a file can carry; a PNG text chunk, for instance, is never listed here even though the re-encode drops it. That is why the cleaned file is read back after every run and the result printed on the page: it is a measurement of the file you now have, not a promise about it.

Because the PDF engine on this site is pdf-lib, which does not decrypt. A protected document fails to load, and the message says so and points you at Unlock PDF, which runs qpdf's decrypt: it removes a password you know, and it clears an owner-password restriction without one. It cannot crack a password you do not have. Run the file through there first and bring the result back here. Images are not encrypted this way, so the image path never runs into it.

No. Transmute processes everything locally in your browser using JavaScript and WebAssembly. Your files never leave your device — there is no server, no upload, no cloud processing.