Image metadata gets all the attention, but audio and video files quietly carry just as much — sometimes more. A video is, after all, a sequence of images plus sound, wrapped in a container that has its own metadata layer on top. When you share a clip or a track, you are potentially handing over where it was recorded, on what device, when, by whom, and occasionally who paid for it.
This guide walks through the main audio and video formats, what each one stores, and the real risks that follow. Every format mentioned here can be cleaned entirely in your browser with our tools — nothing is uploaded.
Video: the GPS problem nobody mentions
When you record a video on a smartphone, the file does not just contain the footage. The container — usually MP4 (.mp4) or QuickTime (.mov) — wraps the audio and video streams in a tree of "atoms" (also called boxes). One of those atoms, udta (user data), and a related meta atom, hold descriptive metadata. On phone recordings this routinely includes:
- GPS coordinates stored in a location atom (often written as
©xyz), pinning where you filmed - The device make and model
- The software version that produced the file
- Creation and modification timestamps
The location data is the dangerous one. People are now broadly aware that a photo can reveal where it was taken, but the same person who carefully strips a photo will happily post a video filmed in their living room — coordinates attached. Because the metadata sits in the container rather than the visible frames, nothing on screen hints that it is there.
How the MP4/MOV container is structured
MP4 and MOV share the same underlying format, called the ISO Base Media File Format. It is a nested structure of atoms, each with a size and a four-character type. Simplified, a phone video looks like this:
ftyp (file type / brand)
moov (movie metadata container)
├── mvhd (movie header)
├── trak (a track: video or audio)
└── udta (USER DATA — GPS, device, etc.)
└── ©xyz (+40.7128-074.0060/)
mdat (the actual audio/video data)To remove the metadata cleanly, you do not re-encode the video — that would be slow and would degrade quality. Instead you walk the atom tree, drop the udta and meta atoms wholesale, and fix the size fields of their parent containers so the file stays valid. The audio and video streams in mdat are never touched, so cleaning is lossless and near-instant.
MP3: the tags that remember everything
MP3 files store metadata in ID3 tags. There are two versions, and a single file can carry both: ID3v2 sits at the front of the file and ID3v1 occupies the last 128 bytes. Between them they hold the title, artist, album, year, genre, track number, comments, embedded cover art, and — in files purchased from some stores — the name or email of the account that bought the track, embedded as a kind of watermark.
That purchaser information is the surprise. People share music files assuming they are anonymous audio, not realizing the file may literally contain their name. Even setting that aside, the comment and encoder fields often reveal which software ripped or edited the file and when.
FLAC, OGG, and WAV
Lossless and open formats have their own metadata conventions:
- FLAC uses a VORBIS_COMMENT block for tags (artist, album, and arbitrary custom fields) and a PICTURE block for embedded album art. Both are separate metadata blocks that can be removed without touching the audio stream.
- OGG (Vorbis or Opus) also uses Vorbis comments, embedded in the bitstream pages. These hold the same kinds of fields as FLAC.
- WAV, despite its reputation as a "raw" format, can carry a LIST/INFO chunk with artist and comment data, an embedded ID3 chunk, and a broadcast extension (
bext) chunk used in professional audio that records the originator, origination date, and time.
Where this actually bites
The recurring theme is that the medium feels anonymous while the metadata is not. A few concrete patterns:
- The activist's protest footage. Video filmed at a demonstration and shared for documentation carries the GPS of the filming location and the phone's identifiers — useful to anyone trying to identify who was present.
- The journalist's source recording. An audio interview saved and shared can carry device and timestamp metadata that helps narrow down where and when it was recorded.
- The purchased track re-shared. A music file passed between friends still names the original purchaser, turning a casual share into an attributable one.
- The professional WAV with a bext chunk. Audio handed to a client carries originator and date metadata that reveals the studio's internal workflow.
What to do about it
For each format, the fix is to strip the metadata layer while leaving the actual media intact. Our engine does this in the browser: it removes ID3 tags from MP3s, the udta and meta atoms (including GPS) from MP4/MOV/M4A, the comment and picture blocks from FLAC, the Vorbis comments from OGG, and the LIST/bext/ID3 chunks from WAV. None of these operations re-encode the media, so there is no quality loss and the file stays playable everywhere it played before.
The practical rule is simple: before you upload a video or share an audio file with anyone outside your circle of trust, run it through a metadata cleaner the same way you would a photo. The location data in a phone video is every bit as revealing as the GPS in a photo — it just hides one layer deeper.
You can strip metadata from images, audio, and video entirely in your browser using the image tool on our home page and the dedicated format pages — no upload, no signup, nothing leaves your device.