local only · files never leave your browser

Batch file metadata scrubber

Drop in images, PDFs, Office documents and text files together. Each one is parsed in memory and rewritten without its provenance metadata — EXIF, XMP, C2PA Content Credentials, PDF info dictionaries and OOXML document properties — then handed back as a single ZIP with an audit report.

Drag files or a folder selection here — PNG, JPG, WebP, GIF, PDF, DOCX, XLSX, PPTX, TXT, MD, CSV, JSON.

What gets removed, by format

  • PNG — tEXt, zTXt and iTXt text chunks (including XMP), the eXIf block, tIME timestamps and caBX C2PA manifests. Pixel data is copied through untouched, so the rewrite is lossless.
  • JPEG — APP1 (EXIF and XMP), APP11 (JUMBF/C2PA Content Credentials), APP13 (Photoshop IPTC) and COM comments. There is no recompression: the scan data is copied verbatim and the ICC colour profile is kept.
  • WebP, GIF, BMP, AVIF — decoded to pixels and re-encoded as PNG, which drops every ancillary block by construction.
  • PDF — the document info dictionary (Title, Author, Producer, Creator, timestamps) and the XMP /Metadata stream where tool provenance and C2PA assertions live.
  • DOCX, XLSX, PPTX — docProps/core.xml, docProps/app.xml and docProps/custom.xml inside the package, plus any Content Credentials parts.
  • TXT, MD, CSV, JSON — the same Unicode provenance cleaning used by the text cleaner.

What this cannot remove

Visible watermark graphics burned into the pixels of an image are not metadata and are not touched here. Neither are statistical token-sampling watermarks such as SynthID-class signals, which live in the content itself rather than in a container header. See the FAQ for the full breakdown.