Remove Content Credentials from images and videos

AI Info is Meta's disclosure label for content that its systems or the creator identify as generated or edited with AI. Meta adopted the wording after the earlier “Made with AI” label felt too absolute on images that had only minor AI-assisted edits.

The label does not tell you which exact signal caused it. A creator can disclose AI use, a file can carry an industry indicator or Content Credential, and Meta can use its own systems. Inspecting the final file helps with the file-level part of that picture.

Start with the final image or video you own, process the file before upload, download the result, and review it. Then use that copy for a new post when the original label came from an accidental supported signal.

A cleaner cannot rewrite media that Instagram already stored. It also cannot override a disclosure that Instagram rules require for the content itself. The dedicated Instagram post guide covers this workflow in more detail.

Use the exact Story image or video you plan to publish. Process that asset before upload, check the output, and keep your source. Story duration does not remove the upload scan from the workflow.

If a published Story already carries AI Info, a web tool cannot edit the file inside Instagram. Return to the source, prepare a new copy, and repost only when that action fits your content plan and the platform rules.

Check the video and the selected cover image as separate assets. Either can have its own provenance or editor data. Use the final versions, not an earlier draft that will be exported again before posting.

Keep disclosure when the Reel is fully generated or materially changed with AI in a way that Instagram asks viewers to understand. File cleanup is for accidental signals, privacy and correct asset preparation.

TikTok says no: once a post receives an automatic AI-generated label, the creator cannot remove that label from the published post. That makes the pre-upload check important when the issue is accidental file provenance.

TikTok also requires labels for certain realistic AI-generated or significantly edited content. Do not use file processing as a substitute for a required creator disclosure.

TikTok names creator disclosure, AI effects, automatic identification and attached Content Credentials among the relevant paths. A video can also violate a policy regardless of whether it has a label.

RemoveAILabel addresses supported signals in the final upload asset. It does not control TikTok's classifiers, creator settings, moderation or policy enforcement.

Google describes SynthID as an imperceptible watermark embedded into supported generated content. For images and video, the signal is part of the media rather than a simple text field in the file header.

C2PA Content Credentials are signed provenance records associated with an asset. They can describe origin, editing actions, tools and AI use. A file can have one, both or neither, so a good report keeps them separate.

No. An AI metadata remover handles stored fields or credentials such as EXIF, XMP, IPTC, PNG text or some C2PA data. A SynthID remover targets a different kind of signal embedded into supported generated media.

Many competing pages use “AI metadata” as a catch-all phrase. This page uses it for searchable language, then explains the technical boundary so visitors know what they are asking the tool to do.

For images, this tool recreates the visible pixels without original embedded metadata. PNG output is lossless, while JPEG and WebP are re-encoded at high quality and can have small compression differences.

For MP4-family video, only the C2PA container boxes are cleared; encoded audio and video tracks are not re-encoded. Keep the original and review every downloaded copy before production use.

No. A platform can use creator disclosure, provenance, an embedded watermark, visual or audio classifiers, account context and policy rules. Those systems can change without notice.

A responsible claim is narrower: the tool can inspect and process the signal types it supports in the submitted file. It cannot promise what every platform or future detector will decide.

The main C2PA remover accepts PNG, JPG and WebP images plus MP4, MOV and M4V videos up to 50 MB. Images also have a 50-megapixel limit. Start with the final full file; cloud placeholders and partial downloads may not contain the asset you expect.

The live format and size limits are shown next to the upload control. Pixel-level and frame-level watermark removal remain separate from this browser-local C2PA workflow.

Only when you have permission to process the file. Keep the client's original, the processed result, and a short record of what was changed. Avoid telling a client that a tool guarantees invisibility to every platform.

Disclosure still depends on the content and destination. If the asset is fully generated or significantly altered in a way a platform requires users to know, keep the label or creator disclosure.

A file tool cannot reach into a live Instagram, TikTok or YouTube post and rewrite the stored media. If accidental file metadata caused the problem, return to the original, prepare a new output copy, and create a new upload when appropriate.

Do not delete a successful post based on a guarantee from an outside tool. Review the platform help page, the content's disclosure needs and the cost of reposting first.