Best fit
Where this guide helps
- Projects mixing generated, live-action, stock, dubbing and post assets
- Teams disclosing AI involvement to platforms, rights holders or clients
- Productions that must reconstruct trailers, episodes and platform variants
Production guide · Provenance & delivery
C2PA 2.4 and AI master delivery: carry provenance, edits and signatures through the pipeline
Separate Content Credentials, ordinary metadata, file hashes and platform AI labels to build a master package that remains auditable through editing and distribution.
Known limits
Where another approach is needed
- Treating C2PA as automatic proof of truth or ownership
- Adding one AI note only at final upload
01
Public documentation
What is known so far
- 01
C2PA 2.4 uses verifiable manifests to record provenance and actions, with signatures and content bindings providing tamper evidence; trust still depends on signer identity and validation context.
- 02
Version 2.4 introduces the crJSON derived view; the specification states that it is not independently verifiable and is not an input format, so exported JSON alone is not the full credential.
- 03
China's synthetic-content labelling rules and platform labels such as TikTok's sit in the publication-compliance layer. They relate to, but are not equivalent to, C2PA credentials.
02
Project setup
Have these inputs ready
- Source, rights, license and file hash for every original asset
- Generation tool, model version, job ID, prompts and reference hashes
- Edit, grade, audio, caption, translation and export actions
- Signer, credential, timestamp and validation policy
- Platform AI labels, end-credit disclosure and territory rules
03
In production
Run the workflow
- 01
Register source, rights, hash and permitted transformations at ingest.
- 02
Create a new version for every generation or edit; never overwrite prior files or manifests.
- 03
Write C2PA actions and ingredients where supported and retain an external audit ledger elsewhere.
- 04
Hash the master after export and validate credentials and signature state.
- 05
Preserve parent-child relationships for platform variants and apply territory/platform labels separately.
- 06
Archive masters, manifests, validation results, disclosures and revision records.
04
Handoff and QC
What the handoff must contain
Deliverables
- Master and platform-variant hashes
- C2PA manifest and validation report
- External asset/action audit ledger
- AI-use and rights disclosure
- Signer, credential and timestamp custody record
QC checklist
- Final file matches recorded hash
- Credential state is distinguished as well-formed, valid or trusted
- Signer and credential status are explainable
- Transcodes and platform variants preserve lineage
- C2PA, end-credit disclosure and platform labels are completed separately
Open questions
What this guide still cannot answer
C2PA supports verifiable provenance and action claims; it does not automatically prove truth, legality, originality or lack of bias. A hash proves file identity only. AniVerse does not treat a credential as sufficient proof that content is trustworthy.