Video metadata remover

What a phone writes into your video

A clip recorded on a phone typically carries a QuickTime location atom holding your coordinates in ISO 6709 format, creation and modification timestamps in the movie and track headers, the device make and model, and the name of the encoder. Editing apps then add their own title, comment and artist fields.

None of it is visible when the video plays, and all of it travels with the file when you send it.

Why we do not re-encode

The obvious way to strip video metadata is to run it through a transcoder. That is what tools built on ffmpeg.wasm do, and it has three costs: it is slow, it downloads tens of megabytes of WebAssembly, and it re-compresses your video so the output is measurably worse than the input.

The container makes a better approach possible. In MP4 and MOV, the stco table inside moov holds absolute byte offsets pointing into the mdat box where the frames live. Deleting a metadata box would shift every one of those offsets and break the file. But the format defines a free box: padding that every demuxer skips. So we rewrite each metadata box's type to free, zero its contents, and leave its size untouched. Nothing moves. Nothing is decoded.

The result is a file the same length as the original, with identical frames, and no metadata.

Streaming, not loading

Browsers will happily run out of memory if you ask them to hold a 2 GB file. We never do. The top-level boxes are walked by reading 16-byte headers through File.slice(), only the moov box is read in full, and the output is assembled as a list of lazy references to ranges of the original file. Peak memory stays in the low megabytes regardless of how long your video is.