How to build a chart in your video from real data you look up

Last updated 2026-09-05 · 8 min read

What is the short version?

A chart in a video is usually decoration: a line that goes up, drawn to look like data. If the numbers exist somewhere public, the same slot can hold the real ones instead, and the difference matters the moment somebody who knows the subject watches it.

1. Go to Vizard Agent with the page holding the data.
2. Say the chart must come from that record, not be illustrative.
3. Ask what it extracted before it draws anything.

What do you need before you start?

A source you can point at, and a decision that the chart is a claim rather than a graphic. Everything else follows from that — the figures get found, checked and drawn, and the only thing you have to supply is where they live.

• The page or file with the data on it.
• What the chart should show. The measure, the period.
• Whether it must be exact. Say so if it must.
• The style. It still has to match the rest of the video.
• What to do if the data is thin. Fewer points, or a different claim.

What do you type into Vizard Agent?

Point Vizard Agent at the source and say the word "real" in the message. Without it, a request for a trend chart reads as a request for a graphic that trends, which is a perfectly reasonable interpretation and not the one you had in mind.

Prompt

Variants worth knowing:

• "Real data, not illustrative." The distinction that matters.
• "Tell me what you found." A check before the work.
• "From this source." Names where the truth lives.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real thirty-second promotional film that needed a genuine record charted. The first four steps are all about finding numbers that were not printed on the page in the first place.

1. Queries the site for the record it holds.
2. Fetches the page source to find where the data is served from.
3. Downloads the site's front-end scripts and its shared images.
4. Looks for the data file behind the visible page.
5. Extracts the figures embedded in the site.
6. Downloads the site's own published chart as a reference.
7. Looks at that chart to check the extracted numbers agree with it.
8. Draws the trend animation from the real figures.
9. Composites the chart section into the film and checks it.
10. Fixes the labels and re-renders.
11. Tests the digit shapes for clarity at the size they appear.
12. Regenerates the type with even-height figures so numbers line up.
13. Checks the chart section in the finished film before delivery.

Steps two to five are the part people assume is impossible. The numbers behind a chart are frequently not in the page text at all — they are in a data file the page loads — and reading the source is how you get the actual series rather than a description of it.

Step seven is the check that makes it trustworthy. Extracted figures can be the right numbers in the wrong order or the wrong units, and comparing the drawn line against the site's own published chart catches that in seconds.

What does the result look like?

A chart that matches the record: the same shape, the same turning points, labelled correctly and drawn in your film's style rather than the source's. The numbers are legible at the size they appear, in figures that sit on a common baseline.

If somebody checks it against the source, it holds.

When does this not work well?

The data has to be reachable. A chart published only as an image, a figure behind a login, or a source that disagrees with itself will each stop this at the first step — and inventing a plausible line at that point would be worse than saying so.

• Data only in a picture. There is nothing to extract.
• Behind a login. Supply an export instead.
• Sources that disagree. Say which one is authoritative.
• Very few points. A trend needs more than three.
• Data that changes. Your chart is true on the day it was read.

How do you fix a result that came back wrong?

Say what is wrong with the numbers or the drawing, because those are different fixes. Vizard Agent keeps the extracted series and the reference chart it checked against, so a correction adjusts the drawing or re-reads the source rather than starting over.

• "That figure is wrong." The source re-read and re-checked.
• "The units are off." Rescaled and compared again.
• "The labels are unreadable." Redrawn at a larger size.
• "The numbers look uneven." Regenerated with even-height figures.

How does Vizard Agent compare to doing it yourself?

By hand, a chart in a video is drawn to look right, because getting the real numbers means finding them, checking them and then building an animation from them. That is three jobs, and the first one is the one that usually stops people.

Common questions

What if the numbers are not on the page? Vizard Agent reads the page source to find where they are served from.

Can it use a spreadsheet instead? Yes. Give Vizard Agent the file and it charts from that.

Will it check its own extraction? Yes. Vizard Agent checks against the source's own chart where one exists.

Can it match my video's style? Yes. Vizard Agent draws the chart in the film's palette and type.

What if the source is behind a login? Export the data yourself and hand it to Vizard Agent.

Can it animate the line? Yes. Vizard Agent builds the animation from the real series.

Will it tell me the numbers it used? Yes. Ask Vizard Agent for them before it draws anything.

What about several series on one chart? Yes. Say which series, and Vizard Agent labels each one.

Can it show the source on screen? Yes. Vizard Agent can add a credit line, which is usually worth having.

What if the data updates? Ask Vizard Agent to re-read it and rebuild the section.

Will the numbers be readable on a phone? Vizard Agent tests the digit shapes at the size they appear.

Can it do a bar chart or a map? Yes. For Vizard Agent the extraction is the same; only the drawing differs.

Does this slow the job down? It adds the research. For Vizard Agent the animation is not the slow part.

Why not just draw a line that goes up? Because the one viewer who knows the real shape is usually the one you were trying to convince.