Metadata stripping: what your files reveal

What is inside a file

Three kinds of data sit around the media itself:

• Container tags. In MP4/MOV: encoder, creation and modification time, device or handler, sometimes GPS. Written by the camera or the exporter.
• EXIF. Mostly photos: make, model, lens, exposure, orientation, capture time, often GPS.
• IPTC and XMP. Rights and editorial data: creator, credit, copyright, keywords, and sometimes an edit history written by the software.

Every re-export writes a new container header, which is why two files from the same batch share an encoder string and a suspiciously regular timestamp.

Why it matters

Privacy

A photo taken at home can carry GPS coordinates. A video can carry the device model and the exact time of capture. Neither is visible in the frame, and both can be read by anyone who downloads the file.

Provenance

Platforms and researchers use metadata as one signal among many. A folder of images all captured at the same second on the same device does not look like a folder of phone photos, however different the pixels are.

Rights and evidence

Credit lines and copyright fields can support a claim — and removing them can weaken one. If you are publishing editorial or licensed work, the licence may require attribution to stay in the file.

What stripping does

Stripping removes the tags rather than editing them. The pixels or samples are untouched; the description around them disappears. For a privacy-first upload, that is often enough: no GPS, no device, no capture time.

What it does not do is change the visual or audio fingerprint. A perfectly clean file is still the same frames and the same soundtrack. Stripping is one layer, not a uniqueness strategy.

Rewriting instead of removing

For image batches, a camera photo with no EXIF at all can look as odd as one with the wrong EXIF. The middle path is to remove the original block and write a new one: make, model, software version and a capture time generated per file. That is what the image uniquifier does, and it is why two images from one batch do not share an identical metadata block.

For video, the equivalent is a clean container written by the encoder with fresh creation data, rather than copying the source header. Some tools re-mux and keep the old tags; that is a half-job.

How to check a file

1. Use a metadata reader. ExifTool is the reference tool; many systems also show basic properties in the file inspector.
2. Look for GPS first. If coordinates are present, decide before publishing.
3. Compare two outputs. If every file in a batch has the same timestamp and encoder, your tool is not rebuilding metadata.
4. Re-check after upload. Some platforms strip metadata on ingest, some keep it, some expose it in downloads. Do not assume.

When not to strip

• Licensed editorial work where the credit must remain attached.
• Legal or insurance evidence, where the original file and its metadata are the record.
• Client delivery where the contract specifies embedded copyright.
• Archival masters. Keep an untouched original; strip only the published copy.

The safe pattern is to keep a master with metadata and produce derivatives without it. Deleting the only copy is not a privacy strategy; it is data loss.

At batch scale

Metadata is per file, so a batch tool has to make a decision per output: strip, rewrite, or keep. RepurposeThis strips the source and writes fresh values by default for both video and images; the image pipeline can also turn the rewrite off. The documentation covers the settings, and the glossary defines EXIF, IPTC and XMP.

Short answers

---

Next: nine changes ranked, or the privacy page.