How to make a documentary about a real event without inventing footage

Last updated 2026-09-22 · 8 min read

What is the short version?

True-crime and historical documentaries are made out of other people's worst days. The line that matters, and the one Vizard Agent holds to, is between a picture that illustrates a point being made and a picture a viewer could reasonably mistake for evidence of what happened.

1. Tell Vizard Agent the case is real and name it.
2. Ask for illustrative footage rather than recreations.
3. Ask it to verify every date, name and figure before it goes on screen.

What do you need before you start?

The narration or the script, and a view on the visual language. Vizard Agent can research the story itself, but documentaries in this genre live or die on their facts, and it is worth deciding early whether archive material will be used at all.

• The voiceover. Recorded, or to be written.
• The case. Named, so it can be verified.
• The visual language. Illustrative, archival, graphic.
• What must be on screen. Dates, statistics, a timeline.
• What must not. Recreations, an invented likeness.

What do you type into Vizard Agent?

Say "illustrative" out loud. Without it, a documentary brief is answered with the most vivid possible images, and vivid in this genre means a generated child on a generated street, which is exactly the thing you cannot publish.

Prompt

Variants worth knowing:

• "Clearly illustrative." The rule for every picture.
• "Not recreations of the events." The specific prohibition.
• "Verify the figure before putting it on screen." Numbers age badly.

What does Vizard Agent actually do?

Here is the Vizard Agent sequence on a documentary about a real case, built from eleven supplied voiceover files. Notice where the verification work sits in the order: before the visuals are chosen rather than as a review pass after the film has been assembled.

1. Checks the supplied audio lengths and formats.
2. Works out what the requested style refers to before choosing a visual language.
3. Measures each clip so the sequence can be assembled accurately.
4. Extracts the spoken content to map the visual beats against it.
5. Diagnoses the one audio file that would not decode, rather than dropping it.
6. Checks the still-image and motion-graphics options for maps, timelines and alert interfaces.
7. Chooses a historically responsible treatment for any generated stills.
8. Sources clearly illustrative stock — a child's bike ride, suburban atmosphere, search and response, emergency alerts, roads and signs.
9. Checks authoritative historical sources for the timeline and the facts.
10. Looks for a trustworthy, publicly usable archive image of the person, if one exists.
11. Verifies the official terminology for the on-screen text.
12. Verifies the recovery statistic before deciding whether to include it at all.

Step twelve is the one that separates a documentary from a video essay. A widely repeated statistic is not the same as a current official one, and a number on screen is a claim you are making rather than a number you read somewhere.

Step eight is the visual discipline. Each of those searches is for something that illustrates a section — suburbia, a search, an alert — and none of them is for a depiction of the event itself.

Step ten is the honest version of using archive material. The question is not whether an image exists but whether it can be used: where it came from, what it actually shows, and whether publishing it is fair to the people in it.

What does the result look like?

A documentary that carries the story without pretending to show it: illustrative footage under the narration, graphics for the timeline and the mechanics, on-screen facts that match official sources, and no image a viewer could mistake for a record of what happened.

When does this not work well?

Some stories should not be made this way. If the only way to tell it is to show what happened, if the family has asked for privacy, or if the facts are genuinely disputed, an illustrated documentary becomes a dramatisation with a documentary's authority.

• Stories that need the event on screen. Illustration is not enough.
• Disputed facts. Presenting one version as settled.
• Living private individuals. Identifiable people who did not choose this.
• Very recent cases. The record is still moving.
• Material under restriction. Court or police images you cannot use.

How do you fix a result that came back wrong?

Say which image or which claim is the problem. Vizard Agent can replace a single visual or re-verify one fact without rebuilding the film around it, and in this genre those two corrections are the ones that actually matter to whether it can be published.

• "That shot looks like a recreation." It is replaced with something clearly illustrative.
• "That figure is out of date." It is re-verified and updated or removed.
• "Do not use that photograph." It is dropped and the section rebuilt.
• "The terminology is wrong." The official wording is used throughout.

How does Vizard Agent compare to doing it yourself?

By hand the temptation is obvious: generated recreations are free, fast and dramatic. Channels that use them get the view count and the correction in the comments, and once one image is shown to be invented the whole film is treated as invented.

Common questions

Can it write the narration too? Yes, though a documentary is better with your own voice and your own research.

Will it tell me its sources? Yes. Ask Vizard Agent which sources each fact came from.

Can it use news footage? Only where the rights allow. Vizard Agent will say when they do not.

What about generated images? Vizard Agent uses them for atmosphere and diagrams, never to depict real people or events.

Should the video say what is illustrative? Yes. A line on screen costs nothing, and Vizard Agent will add it.

Can it build a timeline graphic? Yes. Vizard Agent builds it from the verified dates rather than from the script.

What if a fact cannot be verified? Vizard Agent flags it so you can drop it or soften the claim.

Can it match a channel's style? Yes. Give Vizard Agent a reference and it follows the visual language.

How long should this kind of video be? As long as the material supports — Vizard Agent takes the length from the narration.

Can it handle a file that will not decode? Yes. Vizard Agent diagnoses it rather than silently skipping it.

Will it include a victim's photograph? Only if one is publicly usable, and only if you ask for it.

Can it do a series? Yes, with a consistent visual language across episodes.

What about music? A restrained bed. Vizard Agent keeps it under the narration.

Is this slower than making it up? Yes, and that is the point of the genre.