Skip to content
SynthIDX

Share a Clip Without the Metadata Inside It

Video Metadata Remover

Strips C2PA, container tags and generator fields from video files. It does not remove visible logos or the video pixel layer.

Sign in to run. Free credits included.

Clears the C2PA record, the AI container tags and the generator fields in one pass · The video and audio streams are copied through, not re-encoded · Each strip costs 5 credits, and new accounts get 100 free credits for twenty runs

What is video metadata?

A video file is a container wrapped around the footage. An MP4, MOV, WebM or MKV file holds video tracks, audio tracks and subtitle tracks inside a structure of headers. The metadata is the part that describes the file rather than shows it: the software that wrote it, the device or model behind it, the creation time and, in AI-generated clips, the provenance record the generator attached. It travels in the same file but sits outside the frames.

MP4 and MOV are members of the ISO Base Media File Format, a family built from boxes. Each box has a size and a four character type, and boxes nest inside one another. The MP4 Registration Authority maintains the registry of box and atom types at https://mp4ra.org/registered-types/boxes. A player reads the box tree to find the sample table, the tracks and the metadata entries. C2PA stores its signed video manifest in a top level box of that tree, which is why a container level strip can remove the record without touching the coded video.

C2PA is the signed layer. The Content Credentials technical specification at https://spec.c2pa.org/specifications/specifications/2.5/specs/C2PA_Specification.html lists the ISO Base Media File Format among its embeddable formats and describes the manifest store, the claim and the hard binding in Appendix A.5. For a monolithic MP4 the binding can hash the media data box as a unit; for fragmented MP4 it adds chunk specific hashing. A manifest can name the generator, the actions taken, the timestamps and the signer. In a video it sits beside the footage, not inside the picture.

Generator fields are the other half. A model or an editor can write XMP packets, app export keys and region specific AI labels into the container. A clip from a Chinese AIGC service can carry a TC260 AIGC tag, and a tool can write its own key into a QuickTime metadata list. These fields are not standardized into one location, which is why a strip has to know the containers it handles and why one container can keep more than another.

How it works

How to remove metadata from a video

  1. Step 1: Upload your clip

    Drop an MP4, MOV, WebM or MKV file into the panel above, or click to browse. The upload goes from your browser to processing storage, and nothing is published. The original file stays in place until the strip runs.

    Tip

    Start from the export the generator gave you. A re-shared copy often arrives with less metadata than the original.

  2. Step 2: Run the video metadata strip

    The panel runs a single video metadata operation, configured to write a stripped copy, and it costs 5 credits. For MP4 and MOV the job walks the box tree and blanks the provenance boxes and generator entries in place, keeping every box size and byte offset valid, so the coded video and audio are copied rather than re-encoded. For WebM and MKV the job remuxes the streams with a stream copy and drops the tag list and the chapter list. The result is a new file; the upload is never overwritten.

    Tip

    Credits are consumed only when the strip actually runs.

  3. Step 3: Download the clean copy

    Download the result and check it in a player and in a metadata viewer. The footage and the soundtrack should play as they did, and the provenance fields should be gone. If the clip is going to a platform, clean it at the last step before upload, because a platform can write a new record into the copy it stores.

    Tip

    Keep the original until you have confirmed the export plays and carries no provenance record.

The video metadata this page clears

  • C2PA / Content Credentials

    The C2PA manifest store is the signed record that names the generator, the actions and the signer. In MP4 and MOV it lives in a top level uuid or JUMBF box. The strip turns the box into a free box and zeroes its payload, so a validator that looks for Content Credentials finds none. The C2PA specification describes the embedding rules in Appendix A.5.

  • Container tags

    A container carries AI related tags as well as the ordinary title and author fields. The strip targets the provenance bearing entries: AI label XMP, TC260 AIGC tags, app export keys and the generator metadata list inside the movie box. On WebM and MKV the stream copy drops the whole tag and chapter list. On MP4 and MOV the walker blanks the provenance entries in place and leaves an ordinary title or author field alone when it carries no provenance.

  • Generator fields

    Generators write their own names, model names, prompts and app specific keys into the container. Those fields are what a platform or a verifier reads when it decides whether a clip was made by AI. The strip clears the generator entries it can identify in the container and the app export keys that point at the creating tool. A signal written into the coded stream itself is a separate layer, covered in the limitations below.

What stays after a video metadata strip

A metadata strip changes the record around the footage, not the frames. A visible logo burned into a corner stays in the picture. The video pixel layer, including an invisible signal such as video SynthID, stays as well. The strip has no decoder for that layer and makes no claim about it. If a frame shows a logo, or if a detector reads a pattern in the pixels, the exported file still carries it.

The coded video and audio are copied through. For MP4 and MOV the operation keeps every box length so the sample table and the absolute media offsets remain valid, and the coded content is not re-encoded. For WebM and MKV the stream copy keeps the codec data and drops the container tags and chapters. Either way, the picture and the soundtrack are the ones you uploaded. A metadata strip is not a cleanup of the image itself.

Platforms look at more than the file. YouTube’s help center at https://support.google.com/youtube/answer/14328491 says it may apply an AI label to content that contains C2PA metadata, and it also says a platform can label a video from its own systems or from the creator’s disclosure. Removing the record removes one input. It does not remove a classifier, a manual review or a label the platform has already attached.

Limitations and honest expectations

Visible logos and the video pixel layer stay

This page strips metadata, and it does not remove what lives in the picture. A visible logo burned into the frame stays, and video SynthID and similar invisible signals live in the pixels, so they stay too. No public decoder exists for video SynthID, and this tool makes no attempt to remove it. The AI Video Watermark Remover page is the starting point when a visible mark is the problem.

Metadata inside the coded stream is a separate layer

Some generators write a signal into the video bitstream rather than into a container box. The strip can blank a known tag such as the Kling SEI message, but an unrecognized bitstream payload is outside a container metadata pass, and this page does not promise to remove it. Treat a clean container as one layer removed, not as a guarantee about every artifact in the file.

Platforms can write a new record after upload

A platform that re-encodes or republishes your upload can add its own metadata after the file leaves your hands. YouTube documents two related cases: it may apply an AI label to content that contains C2PA metadata, and it carries forward AI disclosures from C2PA 2.1 or higher credentials that say the whole video was made with AI (https://support.google.com/youtube/answer/15447836). Clean the file at the last step, and check what the destination does with it.

Ordinary tags can survive on MP4 and MOV

On MP4 and MOV the strip blanks provenance entries in place. A title, an author or a creation time field that carries no AI provenance can stay, because removing it would move box offsets that the sample table depends on. If you need every standard tag gone as well, use a general editor and check its export.

Supported containers and unreadable files

The panel accepts MP4, MOV, WebM and MKV. Other containers are not wired into this operation. A file that no parser can open may be copied through unchanged, and a job that finds nothing to strip still produces a copy. Check the result rather than assuming a clean report means every field was removed.

Stripping provenance is a decision about authorship

Authorship and edit history are what C2PA records are for. Strip them from a clip you shot or edited, and nothing is misrepresented. Strip them from a clip that belongs to somebody else, though, and the file now claims a provenance it never had, which can break that party’s terms and, in some places, the law. Keep the pass to material you created or are licensed to change.

Identity and official documents are not supported

This tool is not for passports, identity cards, driver licenses, visas, certificates, statements, or other official documents, whether the document is real, generated by AI, or put together from a template. Editing one of these can be document fraud, and we close accounts that try. Use the tool on other images you generated or have the right to edit.

Where to go next

Which page fits depends on the layer you want gone. For EXIF, GPS and personal fields in a phone photo or a clip export, the Metadata Remover page handles the privacy job. For a visible mark burned into the frames, the AI Video Watermark Remover runs a frame level pass over supported marks. For the signed credential standard itself, the C2PA Remover page covers what Content Credentials record and how a strip removes them.

The facts on this page come from official documentation. The C2PA technical specification is at https://spec.c2pa.org/specifications/specifications/2.5/specs/C2PA_Specification.html. The MP4 Registration Authority lists box and atom types at https://mp4ra.org/registered-types/boxes. YouTube’s help center at https://support.google.com/youtube/answer/14328491 documents its disclosure rules and the C2PA metadata that can trigger an automatic label, and https://support.google.com/youtube/answer/15447836 describes how Content Credentials are carried into its disclosures.

FAQ

Frequently asked questions

Short answers about this video metadata remover, from C2PA boxes to the pixel layer it leaves alone.

Explore more tools

Metadata Remover

The privacy page for EXIF, GPS and device fields in your own photos and clips, stripped from the same metadata layer.

Open tool

AI Video Watermark Remover

For a visible mark burned into the frames: a frame level pass over supported AI video watermarks, separate from the metadata layer.

Open tool

C2PA Remover

The credential specialist: what Content Credentials record, where they live in a file and how a strip removes them.

Open tool

Sora Watermark Remover

The visible Sora logo is outside this metadata pass; the dedicated page tracks it frame by frame.

Open tool

Veo Watermark Remover

Veo clips follow the same split: container tags here, the visible mark on the dedicated page.

Open tool

All tools

Every remover and detector on one page, grouped by what each one cleans or checks.

Open tool

Strip the video metadata before you share

One metadata pass clears the C2PA record, the AI container tags and the generator fields. The footage and the soundtrack come back as they went in.