Define a series question and episode boundaries
A documentary-style series needs a central question broad enough to sustain several episodes and narrow enough to research responsibly. Divide it by chronology, mechanism, place, person, or competing explanation. Each episode should answer one complete sub-question rather than function as an advertisement for a future answer.
Create a series bible with terminology, timeline, disputed points, source hierarchy, pronunciation, recurring visual rules, and a list of scenes that require an “illustration” or “reconstruction” label. This prevents later episodes from contradicting earlier ones without explanation.
Maintain a claim and asset ledger
For every narrated factual claim, record the source, retrieval date, exact supporting passage, confidence, and reviewer. For every image, clip, quotation, and music track, record origin, rights, modifications, and required credit. AI output does not remove copyright, privacy, publicity, or misrepresentation responsibilities.
Separate sourced fact, informed interpretation, and visual reconstruction in the script. A synthetic image of an undocumented event should not be framed as archival evidence. If a dispute cannot be explained within the episode, narrow the claim or defer it.
- Claim ID, wording, source, date, and reviewer
- Asset origin, license, credit, and transformation notes
- Fact, interpretation, and reconstruction labels
- Correction owner and affected-episode inventory
Produce one episode at a time
Complete the research and review loop for the pilot episode before generating the rest of the series. The first output may reveal that the visual approach cannot accurately represent the period, that narration is mispronouncing key names, or that the topic requires more context than short form permits.
Use VidiPrompt for the short-form draft, then verify every scene against the ledger. For a separate long-form compilation, use an appropriate editing workflow and re-check pacing, rights, transitions, and context; do not imply that a set of Shorts automatically becomes a coherent documentary.
VidiPrompt workflow example: a three-part series on an abandoned railway
Episodes cover why it was built, how it changed the region, and why service ended. Reconstructed scenes must be labeled and no motive is assigned without evidence.
- 1Build a timeline and claim ledger from archives, transport records, and reliable local history sources.
- 2Write three self-contained episode questions and define the evidence each requires.
- 3Generate episode one in VidiPrompt with a consistent map, palette, and caption treatment.
- 4Reject inaccurate infrastructure, dates, signage, and unlabeled synthetic “archive” scenes.
- 5Update the series bible from the pilot before producing the next episode.
Review questions
- Can every factual sentence be found in the claim ledger?
- Are reconstruction and uncertainty visible to the viewer?
- Does the episode answer its own question without requiring a later part?
Pre-publish checklist
- Series question and episode map
- Claim, source, and asset ledger
- Visible reconstruction and uncertainty labels
- Pilot episode reviewed before batch production
- Cross-episode correction inventory
Common failure modes
- AI summaries replace reading the underlying sources.
- Synthetic scenes are styled as archival evidence.
- One episode withholds its answer solely to force the next view.
- A collection of Shorts is called a long-form documentary without additional editing and context.
Primary sources
- Disclosing use of altered or synthetic content
YouTube Help — Current YouTube disclosure requirements for realistic altered or synthetic media.
- Copyright on YouTube
YouTube Help — YouTube’s first-party introduction to copyright responsibilities.