Define what the film will explain
Start with a question narrow enough to answer. A film about how people moved materials along a river needs a different script from a survey of an entire civilization. Define the place, period and uncertainty before creating imagery.
In LongVid, keep the working title, research notes and selected sources attached to the production. Channel brain can hold shared writing rules, while the production brief holds the specific setting and permitted cast.
Separate evidence from reconstruction
An attractive generated image is not evidence. Mark which scenes illustrate a hypothesis, which explain a documented process and which should use a map, object or source image instead. Avoid presenting invented boundaries, clothing or architecture as verified fact.
Sources & claims gives you a place to record and review claims. It does not automatically fact-check the documentary. You remain responsible for checking sources, handling disagreements and making clear what is interpretation.
Write a sequence that develops the answer
Give each chapter a job: establish the problem, introduce evidence, examine a limitation and explain what changes in the viewer’s understanding. Read the outline before commissioning the full narration.
After the script is reviewed, the scene planner proposes visual beats tied to passages. Keep an image when the same explanation continues; change it when the subject or evidence changes. Use close details as well as wide establishing views.
Choose the right kind of visual for each passage
Use generated scenes for clearly labelled reconstruction, a restrained camera move for a still detail, and reviewed stock when real footage accurately represents the subject. A stock target percentage is a planning preference, not a reason to insert irrelevant clips.
Recurring figures can have saved reference portraits and fixed appearance descriptions. Review their faces, clothing, scale and setting in every scene. A presenter performance also needs matching narration; identity and lip sync are not guaranteed by the software.
Listen, edit and inspect the rendered result
Test the narrator on names and technical terms. Let the measured audio guide scene duration. In Editor, adjust framing, pacing, chapter titles and captions, then watch the sequence without stopping.
MP4 export currently requires the operator render setup. Invited users can prepare and edit but cannot yet use self-service cloud export. The public demo shows editable illustrative scenes, not a verified end-to-end documentary benchmark.