Veizik can support machine-readable provenance metadata for generated media where the selected workflow and output format provide it.
Provenance metadata is written into the output file or an associated manifest — it states that a machine produced the media and, where applicable, which model. It does not alter the image or video content itself; there is no visible mark on the picture.
Supported output formats are documented per runtime release. Not every export path carries a manifest today.
Machine-readable provenance metadata is enabled across all tiers by current product policy — it is not a paid feature and not a visible watermark.
The provenance claim (machine-generated, model identity where available) is cryptographically signed at write time.
No prompt text, no personally identifying data, and nothing that was not already part of the render configuration.
Provenance-signed files can be inspected with standard C2PA tooling, e.g. c2patool <file>. A signer that is not in a public CA trust list may show as "unverified" in some validators even when the underlying claim is structurally valid — that is a trust-chain distinction, not a claim that the metadata is wrong.
Stripping provenance metadata does not unlock any capability and is not required for any supported workflow. It may affect downstream platforms that specifically check for content-provenance signals.
Related: Model Licences · Privacy