Video & Audio Metadata Checker

A video lands in a group chat, or a client sends an audio clip and asks where it came from. The first thing worth reading is not the picture or the sound — it is the container the file was wrapped in. Every video and audio file travels inside a container, and that container often keeps labels about how the file was encoded, which software touched it, and when it was made.

This video and audio metadata checker reads those labels with ffprobe, the same tool editors and broadcast engineers use to inspect a file. It does not guess from how the clip looks. It reports what the container says about itself, in plain language, and it is careful about what those labels can and cannot prove.

What this checker reads

When you choose a file, the checker reports four things. First, the container — the wrapper format such as MP4, QuickTime, WebM or Matroska, along with the file's length, size and overall bit rate. Second, thetracks inside it — each video track with its codec, resolution and profile, and each audio track with its codec, sample rate and channel layout. Third, the labels written into the container, such as the software that encoded it, a handler name, a creation time, or a title and artist. Fourth, a quick look for acredential marker that may mean the file carries a signed C2PA record.

Those are useful, specific details. A resolution or a codec can tell you whether a file matches the spec someone claimed it had. A creation-time label can flag a mismatch in a chain of custody. A title or artist field can reveal where a clip was first exported. The checker shows all of it as evidence, not as a verdict.

How it works

This check runs entirely in your browser. The file is read on your device and is never uploaded, so it never leaves your computer. Nothing is sent to a server, and nothing is stored or added to any library. Private files stay private.

The read is also one-way. ffprobe inspects a file without changing it, and this site does not offer any rewriting, re-encoding, or watermark removal. If you are looking for a way to strip labels, this is the wrong place.

What a result does not tell you

The most common mistake is reading an encoder label as proof of AI. Some files record a name such asLavf or Lavc, which means they were wrapped or compressed with FFmpeg-family software. That says how the file was packaged — it says nothing about whether the content was generated, filmed, or edited by a person or a model. The checker calls this out whenever it appears so the label is not over-read.

A container check also cannot recover a prompt, detect Google SynthID, or decide whether media is AI-made. SynthID belongs with Google's official detector. And no result here means "human-made" — container labels can be edited or stripped, and many exports simply never carry a credential.

If the checker finds a credential marker, that is a hint, not a full read. To open and verify a signed C2PA record, use a C2PA verifier that supports your format. For images, the Content Credentials checker does that in the browser. For video, that verification needs a tool that can read the manifest inside the file.

Why container labels go missing

Platforms often re-encode a video or audio file when you upload it, and that process can drop or rewrite the labels that were there before. Messaging apps, social networks, and download tools can all change the container. A file that shows almost nothing today may have carried detailed labels when it was first exported.

That is why the copy you check matters. Ask for the original export when the answer matters, and treat a clean-looking result as a statement about that one copy — not about every version that ever existed. For the bigger picture of what a file can and cannot reveal, start with what can be detected in AI media.