Remove metadata from video. No re-encoding.
Clear AI labels, C2PA Content Credentials, GPS location, device details and encoder names from MP4 and MOV files. The video and audio streams are not touched, and the file never leaves your browser.
What's hiding in an MP4 or MOV file
A video file is a set of nested boxes. Most of them describe how to play the streams, but a handful exist only to describe the file, and those are the ones that identify where it came from.
| Where | What it can contain | Removed |
|---|---|---|
uuid box (C2PA) | Content Credentials manifest, including an AI-generated claim and the signing app | Yes |
uuid box (XMP) | XMP packet: creator tool, AI source type, editing history | Yes |
moov/udta | Title, comment, GPS location (©xyz, loci), camera make and model, software | Yes |
moov/meta, trak/meta | Apple QuickTime keys: location, device, software version | Yes |
hdlr, stsd names | Handler and compressor names that reveal the exporting app | Yes (blanked) |
mvhd, tkhd, mdhd | Creation and modification timestamps | Yes (zeroed) |
| Timed metadata track | Per-frame data such as live location, stored as its own track | Reported, not removed |
mdat | The actual video and audio | Never touched |
Why no re-encoding matters for video
Most "remove metadata from video" advice ends with an ffmpeg command or an upload to a server that re-muxes or re-encodes your file. Re-encoding lowers quality and takes minutes. Even a clean re-mux means sending your footage to someone else's machine.
This tool rewrites only the boxes listed above. Removed boxes become free boxes of the same size with their contents zeroed, so every pointer into the media data stays valid. The output is the same size as the input, players treat it as a normal file, and every video and audio packet keeps the exact same checksum.
Large files
The video data is never loaded into memory. Only the small index box (moov) is read, and the rest of the file is passed through in slices when you download. In our tests a 760 MB file was analysed in under 10 milliseconds with less than 1 MB of extra memory. There is no size cap.
AI video and labels
TikTok reads C2PA Content Credentials to automatically label AI-generated videos uploaded from other platforms (TikTok newsroom), and in an MP4 or MOV the manifest sits in a top-level uuid box, which this tool blanks. Meta has said that for video and audio it relies on people disclosing AI use themselves and may apply penalties when they don't (Meta).
Some AI video models also embed an invisible watermark in the frames. Google applies SynthID to video made with Veo (Google). That watermark is part of the picture, so it survives metadata removal. This tool does not alter the frames and does not try to remove it.
Three ways to remove metadata from a video, tested
We ran the same C2PA-signed MP4 (a test file signed with an "AI-generated" claim) through two approaches and compared the results with ffprobe and the official C2PA library.
| This tool | ffmpeg -map_metadata -1 -c copy | |
|---|---|---|
| C2PA manifest | Removed | Removed |
| XMP packet | Removed | Removed |
| Video and audio packets | Identical | Identical (copied) |
| Container | Original structure kept, same file size | Rewritten: brand changed to isom, about 8% smaller |
| New tags written | None | Its own handler names (VideoHandler, SoundHandler) |
| Language tags | Kept | Reset to und |
| What you need | A browser | ffmpeg installed and a command line |
If you already use ffmpeg, that command is a perfectly good option, and we would rather say so than pretend otherwise. This tool exists for everyone who doesn't want to install anything, for batches of mixed photos and videos, and for cases where you want the container left exactly as it was. The third common approach, exporting again from a video editor, re-encodes the picture, which costs quality and time and usually writes the editor's own tags.
Step by step
- Drop one or more MP4 or MOV files into the box above. Nothing is uploaded; the browser reads the file directly from your disk.
- Read the report. It lists every metadata box it found, with the path inside the file (for example
moov/udta) and what it contains. - Wait for "re-checked: clean". The cleaned file is parsed again from scratch before you are allowed to see that status.
- Download the file, or download several at once as a .zip. Your original file is not modified.
What stays exactly the same
- Codec, resolution, frame rate, bitrate and duration
- Rotation (the display matrix in the track header is kept, so portrait clips stay portrait)
- Audio tracks, subtitles and chapter structure in the media data
- File size: removed boxes are replaced by empty boxes of the same length
Check the result yourself
After cleaning, each file is parsed again from scratch and the report shows what is left. If you have ffmpeg installed you can also run ffprobe -show_format -show_streams file.mp4 on the original and the cleaned copy: the format and stream tags are gone, while the codec, resolution, duration and bitrate are unchanged.
Frequently asked questions
Does removing metadata reduce video quality?
No. The video and audio streams are copied as they are. Only metadata boxes are blanked, which is why the cleaned file is the same size as the original.
Will the cleaned video still play everywhere?
Yes. Blanked boxes are converted to standard free boxes that every player is required to skip. We check every release against ffmpeg and ffprobe.
Can I clean a video from my iPhone?
Yes. iPhone videos are MOV or MP4 files and carry location, device and software tags in QuickTime metadata boxes. Those are removed. Some iPhone clips also contain a separate timed metadata track; the report tells you when one is present.
Does this remove visible watermarks from AI videos?
No. A logo or watermark drawn into the frames is part of the picture, not metadata. This tool only removes data stored alongside the video.
Is there a size limit?
No. Multi-gigabyte files work because the video data is streamed, not loaded into memory.