RemoveAILabel
Cleaning tool · Video

Clean MP4 / MOV video metadata

A video's metadata lives at the container layer: udta (user data) and meta (iTunes tags). Removing them needs no re-encoding — the picture does not change at all.
MP4 · MOV/25 MB per file/5 per batch/No re-encode
Drop a file here, or click to choose
MP4 / MOV only · 25 MB per file · up to 5 per batch
You can also paste a screenshot with ⌘V / Ctrl+V
Cleaning options
Two things to know: first, 25 MB is a hard limit imposed by browser memory — trim long videos in a local editor first. Second, a video's cover image (JPG/PNG) is a separate file that has to be cleaned on its own; that is the usual reason a video is clean while its thumbnail still gets flagged.

Which layer the metadata sits in

LocationContentsAction
moov/udtaUser data: title, author, comment, softwareBlock removed
moov/meta/ilstiTunes-style tags: ©nam, ©too, ©cmtRemoved together with udta
uuid (C2PA)Content credentialsAtom removed
mdatH.264 / H.265 video dataUntouched
Audio and subtitlesAudio samples, subtitle tracksUntouched

Why deleting container tags requires fixing offsets

This is where video cleaning goes wrong most often. An MP4's stco / co64 tables point at video frame positions using absolute byte offsets. If moov sits before mdat — the fast-start layout social platforms require — deleting bytes inside moov shifts mdat forward and every offset becomes wrong. The file still plays, but the picture is garbage.

This tool counts the bytes actually removed and corrects the offset of every frame. That is why it takes the extra step, rather than the lazier route of marking the file dirty without rewriting it.

Video and cover: two separate jobs

  • Video container: handle it on this page, dropping udta / meta.
  • Custom cover: usually a JPG / PNG — run it through the image cleaner.
  • Why it matters: Reels and TikTok use a separate cover file, so cleaning the video alone often leaves the label in place.

Common questions

Will the video be re-encoded?
No. Only the metadata boxes at the container layer are rewritten; the H.264 / H.265 stream and the audio tracks are preserved as they were. Duration, resolution and bitrate are unchanged.
Why the 25 MB limit?
Because the whole file has to be read into memory and rebuilt in the browser; a very large file will simply crash the tab. 25 MB is the compromise between compatibility and usability.
Can ffprobe still read the tags afterwards?
No. udta / meta are removed whole, so common tools no longer show title, comment or encoder. What remains is format identification such as major_brand.
Why did my video fail to process?
Common causes: a corrupt file, an incomplete download, or a non-standard fragmented MP4 layout. Try remuxing once with ffmpeg, or shorten the clip to reduce its size.
Does it support MKV, AVI or WebM?
No. MKV and AVI use entirely different metadata structures, and WebM is a Matroska subset. Only MP4 and MOV are handled.

Related tools