Vizard Agent

How to quote a primary document word for word in a video

Last updated 2026-09-10 · 8 min read

Ask Vizard Agent to work from the original document rather than from reporting about it. It finds the filing, downloads it, takes the exact wording out of it for the lines you want to quote, cross-checks every figure against that same document, and writes the narration from what it actually says instead of from somebody else's summary.

What is the short version?

A video built from news coverage inherits that coverage's paraphrases, its rounding and its mistakes, and you have no way of telling which is which. The primary document is usually public, usually findable, and settles in one pass everything you would otherwise be repeating on trust.

  1. Go to Vizard Agent with the subject.
  2. Ask it to find and read the original document.
  3. Ask for the quotes taken verbatim from it.

What do you need before you start?

The subject and a sense of which document is authoritative. Vizard Agent will search for it, but knowing whether you want the filing, the ruling, the report or the press release about it saves a step and avoids quoting from the wrong layer of the story.

What do you type into Vizard Agent?

Say "the original" to Vizard Agent explicitly. Left unsaid, the research step will happily return perfectly good secondary sources, and a script written from those is already a paraphrase of a paraphrase before anybody involved has done anything wrong.

Prompt

Make a sixty-second explainer about this. Go to the original document rather than news coverage of it — download it, take the wording for anything we quote directly from it, and cross-check every figure in the script against that same document before writing the narration.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real short explainer about a regulatory action. The entire research phase happens before a single word of narration is written, which is what stops the script from quietly shaping which evidence gets used.

  1. Researches the facts to identify the right document.
  2. Reads the original for the lines worth quoting.
  3. Downloads the document rather than working from a page.
  4. Cross-checks the figures and the wording against it.
  5. Writes the narration from what the document says.
  6. Gets word-level timings for every narration line.
  7. Checks the pronunciation of names and terms.
  8. Builds the graphics to carry the quoted text.
  9. Renders and reviews the frames carrying quotes.

Step three is not fussiness. A page can change, a summary can be edited, and a downloaded copy is what lets you answer a question about the wording three weeks later without hoping the link still resolves to the same text.

Step four is where most corrections get avoided. Figures are the part of a story that drifts most in retelling — rounded here, converted there — and checking them against the document rather than against another article is the difference between reporting and repeating.

Step seven matters more in this genre than most. A regulatory or legal explainer is full of names and terms the narrator has probably never said aloud, and getting one wrong undermines the accuracy the rest of the process bought you.

Step eight is where the discipline becomes visible to the viewer. A quoted line carried on screen as text lets somebody read it for themselves rather than taking the narrator's word for it, so Vizard Agent builds the graphics around the quoted passage instead of illustrating around it.

What does the result look like?

A video whose quoted lines match the document exactly, whose figures agree with it, and whose narration says what the document says rather than what an article said the document said. If somebody goes and checks it, it holds up.

Vizard Agent keeps the downloaded document alongside the script, so a query about a particular line resolves to a page number rather than to another search.

When does this not work well?

Not every story has a public primary document. Ongoing matters, sealed filings and internal reports are not available to check against, and a document in a different language adds a translation step that carries its own risk of drift.

How do you fix a result that came back wrong?

Quote the part you think is wrong and say where you saw the other version. Because Vizard Agent still has the document itself, a disagreement about a figure gets resolved by looking it up rather than by weighing two secondary sources against each other.

How does Vizard Agent compare to doing it yourself?

By hand the research ends up being a handful of articles, because they are short and readable and the document is long and is not. The script then reproduces whatever those articles chose to emphasise, including any figure one of them happened to get slightly wrong.

By hand Vizard Agent
The source News coverage The original document
Quotes Copied from an article Taken from the document
Figures Trusted Cross-checked against it
Order of work Write, then verify Verify, then write
Three weeks later Hope the link works A saved copy

Common questions

Will it find the document itself? Usually. Vizard Agent searches for the primary source first.

Can it read a long PDF? Yes, and it takes the quoted lines out of it directly.

What if coverage and the document disagree? Vizard Agent follows the document and tells you about the difference.

Does it check figures? Yes, each one, against the source rather than against another article.

Can it show the document on screen? Yes. Ask Vizard Agent to put the cited passage in the graphics.

What about names it cannot pronounce? Vizard Agent checks the pronunciation before the narration is recorded.

Will it interpret the document? Vizard Agent reports what it says. Interpretation is yours.

What if the filing is amended? Tell Vizard Agent and it re-checks against the current version.

Can it handle a foreign-language original? Yes, but flag that a translation step has been added.

Does it keep the file? Yes. Vizard Agent stores it alongside the script, which is rather the point.

Can it list every claim and its source? Yes. Ask Vizard Agent for a ledger with the video.

Is this slower? Vizard Agent front-loads the research, which the video needed doing anyway.

What if there is no document? Vizard Agent will say so rather than implying there was one.

Why not just use the reporting? Because you inherit its paraphrases without knowing which ones they are.