A Claim Ledger Is a Production Artifact, Not a Research Appendix

A source list at the end of a documentary project is useful. It is not enough.

The film does not publish a bibliography. It publishes sentences, images, captions, causal implications, quotations, and reconstructions in a precise sequence. The production record should follow those claims at the same level.

That record is a claim ledger.

Claims change after research

Research may begin with a broad question and a collection of promising sources. The script then selects, compresses, attributes, and connects that material. Narration makes the wording sound more certain or less certain. Visuals add implications that the sentence did not state.

A claim can become inaccurate without introducing a false source. Two supported facts can be joined by an unsupported causal phrase. A cautious paragraph can become an absolute voiceover line. A realistic image can imply that an undocumented event was photographed.

The ledger needs to travel with those transformations. Each material claim should record its current wording, classification, supporting source, exact locator, uncertainty, reviewer, and decision.

Visuals make claims too

Documentary review often treats images as decoration and narration as the factual layer. That separation fails with realistic generation.

Clothing places a person in a period. Architecture places a scene in a region. A map establishes geography. A newspaper headline appears to document a statement. A familiar face can imply that a specific person was present.

The claim ledger should therefore include visual propositions. It does not need to inventory every cloud or camera move. It does need to identify scenes whose realism can alter what the viewer believes.

This is especially important for AI historical reconstruction, where the visual can feel evidentiary even when it is an illustration.

Corrections need receipts

A reviewer who finds an unsupported date should not only change the script. The production needs to know whether that date already reached narration, subtitles, chapter text, a map, an image prompt, or the final render.

The ledger creates a correction path:

1. identify the atomic claim;
2. record the source problem;
3. revise or remove the wording;
4. list affected production artifacts;
5. regenerate or replace only what changed;
6. verify the accepted export;
7. close the issue with reviewer and date.

This is slower than changing one document and faster than discovering the same error after publication.

The accepted export is the factual object

A correct script does not prove that the final film is correct. The wrong narration version may remain active. Captions may preserve old wording. A realistic shot may contradict the revised line. The final factual review must watch the accepted export against the ledger.

The ledger should then be frozen with the export identifier, source pack, rights record, disclosure decision, and correction log. If the film changes, the factual record changes with it.

That is why a claim ledger is not administrative residue. It is part of the production architecture, like the timeline or audio master.

What Onira can claim today

Onira currently describes its workflow as research-assisted; creator verification is required. A future customer-facing claim-to-source pack would make the ledger more visible and operational, but it must pass end to end before marketing presents it as shipped.

Until then, creators should use the documentary fact-checking workflow, build a bounded documentary source pack, and keep the verification record outside the generated prose.

The standard is not whether an AI system can produce citations. The standard is whether the person approving the film can explain why each material claim and realistic implication belongs in the final cut.