编辑元数据

这款工具的作用

Every PDF carries a set of details about itself: a title, an author, a subject, keywords, the program that produced it and two dates. Readers show them in a properties dialog, some show the title in the window bar instead of the file name, and search tools index them. This tool reads what is in your document, lets you change any of it, and writes the result back without touching a single page.

Use it when the details are wrong or revealing: a document exported from a colleague's template that still carries their name, a report whose title is "Microsoft Word - draft3.docx", a file going outside the organisation with an internal project code in its keywords, or a scan with no details at all that you would like to be findable.

工作原理

1. Drop the PDF onto this page. Every box is filled in with what the document currently holds.
2. Edit the boxes you care about and leave the rest alone.
3. Empty a text box to remove that detail from the file entirely. An empty date box leaves the date as it is.
4. Set the dates in ISO form under advanced options if they need correcting, and clear the XMP packet there if you want only the classic details left.
5. Press Save metadata and the download starts.

The boxes are filled from your document before you touch them, and that is the whole design. An editor that starts empty makes you retype what is already there to change one field, and silently blanks everything you did not retype. Because the boxes are pre-filled here, an empty box can safely mean "remove this" - which is the only way a form that submits every field can express the difference between clearing a value and leaving it alone.

A PDF stores its details twice: the old Info dictionary and the newer XMP packet, and every modern reader believes the packet when the two disagree. Editing only the dictionary is the classic metadata-editor bug - Acrobat keeps showing the old title and you conclude the tool is broken. So when your document already has a packet, it is rebuilt from the details you just saved. That rebuild does discard producer-specific XMP extensions written by Illustrator or InDesign, which is a real loss and a smaller one than showing you a title you had just deleted.

Dates are ISO 8601 because a PDF date is unambiguous and a localised one is not. "03/04/2026" is two different days depending on where the reader lives, and metadata is precisely where that ambiguity outlives everyone involved. Write 2026-03-04, or 2026-03-04T09:30:00Z when the time matters. An empty date box leaves the existing date alone rather than clearing it, which is the one place where the boxes here differ from each other.

The producer field is left exactly as your document had it. This tool does not stamp its own name over whatever wrote the file, because the producer is often the one honest clue about how a document was made, and quietly rewriting it would make every file that passed through here look the same.

If what you want is nothing rather than something else, Remove metadata is the tool: it deletes the objects instead of the references, since pdf-lib writes out every object in the file whether or not anything points at it, and a packet that is merely unreferenced is still there for anyone with a hex editor.