Universal metadata scrubber

PDFs, JPEG, PNG and WebP images, and Office documents in docx, xlsx and pptx. Each goes to the engine the site already has for it: pdf-lib for PDFs, a decode and re-encode for images, and an OOXML property stripper for Office files. Any other extension is refused with a message naming the formats it handles.

Because it has no verified way to do it. The message says video metadata stripping is not implemented, and explains that the available library documents reading tags rather than rewriting them. Refusing is the honest outcome; passing your file through untouched while claiming success would not be.

The core and app property fields, which are creator, last modified by, subject, title, description, keywords, category, company and manager. All custom properties are removed wholesale, since they have no fixed names. And tracked change and comment author names, initials and user ids in the Word parts are set to Author with their timestamps deleted, which covers the people part where Word keeps a commenter name and email address.

The output line names it, and that list was derived by scrubbing real files and grepping the result. Body text is untouched, including anything hidden or inside a tracked change. Comment authorship inside xlsx and pptx parts is not anonymised, only the Word parts are. Edit session ids, the attached template path, created and modified timestamps and the authoring tool fields also stay.

For JPEG, yes, a little. The image is decoded to pixels and re-encoded at quality 92, which is what removes every metadata block there is. PNG and WebP go through the same path. If you need the original pixels untouched, this is not the tool for it.

Images and Office documents come back with a -clean suffix before the extension, and the PDF path returns a file with a -meta suffix, having cleared its title, author, subject, creator, producer and keywords. Run Privacy risk scanner on the result to see what a scrub cannot reach.