How to explain a system of notation with a worked example on screen

Last updated 2026-09-17 · 8 min read

What is the short version?

A presentation that says a system "used knots to record numbers in base ten" tells the viewer nothing they can use. A card that encodes one specific number, position by position, teaches the whole system in twenty seconds.

1. Give Vizard Agent the topic and how long the video should run.
2. Ask for one worked example, carried all the way through.
3. Ask it to mark what is established and what is still debated.

What do you need before you start?

Only the topic, honestly. Vizard Agent researches the material, checks it against references and writes the narration, but telling it the audience and the length changes the depth considerably — a class of undergraduates and a general audience need different amounts of detail.

• The topic. A title is enough to start from.
• The length. Three minutes, five, ten.
• The audience. School, undergraduate, general.
• The language. For narration and captions.
• Any source you must use. A set text, a paper, a syllabus.

What do you type into Vizard Agent?

Ask for the example explicitly, in the first message. Left to a general brief, any explainer drifts towards description, because description is easier to write and it sounds authoritative on the page; naming the worked example in the instruction is what forces the structure that actually teaches somebody the rules.

Prompt

Variants worth knowing:

• "Work one number all the way through." The structural instruction.
• "How zero is handled." The edge case that proves the system.
• "What is still open." Stops it overstating the scholarship.

What does Vizard Agent actually do?

Here is the Vizard Agent sequence on a four-minute presentation about quipu knot types. The research comes before any writing, and the cards are built around one encoded number rather than around the headings of an essay, which is the difference between a lecture and a set of slides read aloud.

1. Verifies the key facts about the knot types before writing anything.
2. Fetches references for the numerical system and the knot forms.
3. Writes the narration and counts it against the target runtime.
4. Generates a voice sample first to measure pacing before the full read.
5. Generates the historical visuals — a close-up, a museum context, an overhead of the cord structure.
6. Builds the cards: title, knot taxonomy, place value, the worked example, zero, known-versus-open, closing.
7. Transcribes its own narration with word timings so the captions land on the word.
8. Measures narration and music loudness before mixing them.
9. Assembles the presentation and inspects a contact sheet of every card.
10. Fixes card timestamps where a diagram was cut short, and re-renders.
11. Lowers and resizes the captions so they stop covering the diagram labels.
12. Corrects the scholars' names in the timed caption source and renders again.
13. Replaces the ending card and holds the knot diagram longer.

Step six is where the teaching happens. The worked example gets its own card, and so does zero, because a system's treatment of nothing is exactly the part a description glosses over and the part a student will be asked about.

Step eleven is a small thing that decides whether the video works at all. Captions sized for a talking head sit straight across the labels of a diagram, and a diagram you cannot read the labels of is decoration. Vizard Agent checks the frames rather than trusting the caption preset.

Step twelve is worth noting for anyone making academic material. Proper names come out of speech-to-text unreliably, so Vizard Agent corrects them in the timed source rather than leaving a misspelt scholar on screen for four seconds.

What does the result look like?

A narrated presentation of the length you asked for, with captions that sit clear of the diagrams, a single example carried through the system on its own card, and an explicit statement of what is established and what remains contested. Vizard Agent reviews the render and its audio balance before delivering it.

When does this not work well?

Some topics have no settled account to present. Where the scholarship is genuinely divided, or where the sources are thin and contradictory, a confident four-minute explainer misrepresents the field, and Vizard Agent will tell you that rather than picking a side and narrating it.

• Contested scholarship. There is no single account to give.
• Very recent research. References may not be stable yet.
• Topics with no visual form. Nothing for the cards to show.
• Strict citation requirements. Ask for a source card, and check it.
• Niche terminology. Speech synthesis mispronounces it without guidance.

How do you fix a result that came back wrong?

Say which card is at fault rather than asking for a new video. The cards are built separately, so Vizard Agent can hold one longer, rebuild another or re-render a caption line without touching the narration you already approved.

• "The example goes too fast." That card is held longer and re-rendered.
• "The captions cover the diagram." They are resized and moved down.
• "That name is spelt wrong." It is corrected in the timed caption source.
• "It states too much as settled." The known-versus-open card is expanded.

How does Vizard Agent compare to doing it yourself?

By hand this is an evening of slides. The research is the honest part of the work and it takes most of that evening; then the narration gets recorded twice because the first pass ran long, and the captions get skipped entirely because nobody has the patience to time them.

Common questions

Do I have to supply the material? No. Vizard Agent researches the topic and verifies the facts before writing.

Can I supply a set text? Yes, and it is better if the course requires one — Vizard Agent will work from it.

How long can the presentation be? Several minutes is comfortable. Vizard Agent counts the script against your runtime.

Will the narration sound like a machine? Vizard Agent samples a voice and measures its pacing before committing to a full read.

Can it use my own voice? Yes, if you record the narration — Vizard Agent times the captions to it.

What about citations? Ask for a source card and Vizard Agent puts the references on screen.

Can it do the video in another language? Yes, narration and captions both.

Why does the worked example matter so much? Because a rule described is forgotten and a rule demonstrated is not.

Will the historical images be accurate? They are illustrations. Vizard Agent keeps the claims in the narration, not in the pictures.

Can I get the script separately? Yes, ask Vizard Agent for the narration text and the timed transcript.

What if a term is mispronounced? Tell Vizard Agent how it should sound and it re-reads that line.

Can students download captions? Yes. The caption file is produced alongside the video.

Is it suitable for assessment? Check the facts yourself before submitting anything — Vizard Agent shows you what it used.

Can it make a shorter version? Yes, from the same cards, which keeps the two consistent.