How to make a video that explains the scale of something
Give Vizard Agent the quantity and the everyday thing you are mapping it onto. It works out the arithmetic, draws every diagram programmatically from those numbers rather than generating a picture of a chart, and cuts them against footage with a timed narration script.
What is the short version?
A number nobody can picture is not information. The fix is a comparison — the solar system on a football pitch, a company's water use in bathtubs — and the whole video stands or falls on whether the arithmetic is right and visible.
- Go to Vizard Agent and give it the quantity you are explaining.
- Say what everyday thing it should be mapped onto.
- Say the length, and whether you will record the narration yourself.
What do you need before you start?
The numbers and the comparison, and not much else. Vizard Agent does the arithmetic and shows its working, but choosing what everyday thing to map the quantity onto is the creative decision that makes the difference between a video people understand and one they forget immediately.
- The quantity. With its units and source.
- The comparison. A pitch, a room, a city block.
- The points to cover. What the video must establish.
- The length. Four to six minutes for a full treatment.
- The narration. Generated, or a script for you to read.
What do you type into Vizard Agent?
Give it the scale factor if you have one, and the list of things the video must establish. Vizard Agent computes the positions from the scale rather than approximating them, so a stated ratio produces a diagram you can defend when someone checks it.
Prompt
Variants worth knowing:
- Ask for exact-maths diagrams. Not a picture of a chart.
- Ask for the cue sheet. So you can narrate it yourself.
- Set the music low. Quiet enough to speak over.
What does Vizard Agent actually do?
Here is the order Vizard Agent worked in on a real scale-model explainer. The decision that shapes everything is step four: it computes the model's arithmetic first and then draws every panel from those figures, which is why no number on screen is a guess.
- Checks the workspace and loads the explainer approach.
- Checks the graphics and stock tools available to it.
- Works out the scale-model arithmetic — positions, sizes, ratios.
- Searches for stock footage and checks which fonts are installed.
- Downloads the clips and reviews their frames.
- Renders the diagram panels from the computed figures.
- Tiles the panels for review and reads them at full size.
- Re-renders corrected panels plus the pattern and reflection cards.
- Tests which fonts carry the symbols used and fixes the exponent and arrow glyphs.
- Builds the title card, fetches music and renders every segment.
- Builds a very quiet music bed and measures it before muxing.
- Samples frames across the render and watches the whole cut back as stills.
- Replaces two weak shots, re-renders them and verifies the replacements.
- Times the script against the visuals, rebuilds the video to those cues, and writes the narration script with cue times.
Step nine is the kind of thing that only shows up when you look. Superscripts and arrows are missing from many typefaces, so Vizard Agent tested the symbols it needed and wrote a dedicated renderer rather than shipping a panel with a box where an exponent should be.
What does the result look like?
The brief this page is written from asked for four to six minutes, and the cut landed at about five minutes. Dark astronomy styling, programmatically drawn panels — yard lines, planet dots, spacing charts — intercut with stock footage, with slow continuous motion across each panel.
The delivery included a timestamped narration script and a music bed mixed deliberately low so the student could record their own voice over it. That is the right shape for an assignment: the arithmetic and the pictures are built, the voice is theirs.
When does this not work well?
A scale video is only as good as the numbers, and the numbers are yours. Vizard Agent computes the model from what you supply and shows its working, but it cannot check that your source figure was right in the first place.
- A wrong input figure. Everything downstream inherits it.
- A comparison nobody knows. Pick something universal.
- Too many points. Five ideas is plenty in five minutes.
- Symbols in odd fonts. Superscripts need checking.
- Generated charts. Ask for computed ones instead.
How do you fix a result that came back wrong?
Give the corrected number, or name the point that is missing from the argument. Vizard Agent keeps the computed model, the panel renderer and the cue timings, so a changed figure redraws only the panels that depend on it and re-times the narration script around them.
- "That distance is wrong." Recomputed and the panels redrawn.
- "Add the asteroid belt." New panel, script re-timed.
- "The music is too loud to talk over." Re-mixed lower.
How does Vizard Agent compare to doing it yourself?
By hand this is a spreadsheet, a drawing tool and a video editor, kept in sync manually — and every corrected figure means redrawing a panel and re-timing the narration. Vizard Agent draws from the computed model, so a corrected number propagates.
| By hand | Vizard Agent | |
|---|---|---|
| The arithmetic | Spreadsheet, retyped into slides | Computed once, drawn from |
| Diagrams | Drawn by eye | Rendered from exact figures |
| A corrected number | Redraw and re-time | Recompute and re-render |
| Narration | Recorded, then re-timed | Cue sheet timed to the visuals |
Common questions
Why not just generate the diagrams? Because generated charts invent numbers. Vizard Agent draws them programmatically from the computed model.
Do I need to supply the scale factor? Not necessarily. Vizard Agent works out a sensible one and explains why it fits.
Can I record my own narration? Yes, and for an assignment you should. Vizard Agent delivers a timestamped script and a quiet music bed.
Will it generate a voice instead? If you ask. Vizard Agent records the narration and times the visuals to the words.
How long should it be? Four to six minutes for a full treatment. Vizard Agent fits about five ideas into that comfortably.
Can it check my numbers? It can research them, but the source is yours. Vizard Agent computes correctly from whatever figure you give it.
What about symbols and exponents? Checked explicitly. Vizard Agent tests which fonts carry the glyphs it needs before rendering.
Can it mix diagrams with real footage? That is the usual structure. Vizard Agent intercuts computed panels with stock shots of the comparison.
Does it work for non-scientific numbers? Yes. Vizard Agent handles budgets, distances, water use or emissions the same way.
Can I get the panels as images? Yes. Vizard Agent renders them as files you can reuse in slides or a document.
How is this different from an animated explainer? An explainer illustrates an idea; this one has to be arithmetically defensible. Vizard Agent computes the model first precisely so that every position on screen survives someone checking it.
What if my teacher wants sources cited? Ask for a source card. Vizard Agent adds one listing where each figure came from.
Can I change the comparison later? Yes. Vizard Agent recomputes the whole model against a new one rather than adjusting the drawings.