Vizard Agent

How to take a product's claims off its packaging instead of from memory

Last updated 2026-09-28 · 8 min read

Tell Vizard Agent to take every stated figure from the packaging in your own photographs. The panel is already in the picture at higher resolution than the finished video needs, so it can be enlarged and read rather than remembered — which is how a wrong number gets onto a closing card in the first place.

What is the short version?

You are making an ad from product photos, and the card at the end states a figure. If that figure came from memory or from a summary, it is a guess. Vizard Agent reads it off the packaging you already sent.

  1. Tell Vizard Agent the packaging is authoritative.
  2. Have it enlarge the panel and read the figures.
  3. Keep the wrapper fully in frame so the claim is checkable.

What do you need before you start?

Photographs where the panel is legible at full resolution. It does not need to be readable in the finished crop — it needs to exist somewhere in the source files at enough detail to be enlarged and read, which most phone photos of a package easily satisfy.

Say what the ad is allowed to claim. A benefits card that names a number is a factual statement about the product, and it is worth deciding in advance which figures go on screen rather than discovering one on the delivered file.

What do you type into Vizard Agent?

Name the packaging as the source of truth, explicitly. Without that instruction, a figure sitting in your brief is the most authoritative thing Vizard Agent has, and a number you typed from memory will be reproduced onto the card faithfully and confidently.

Make a reel for this protein bar from these photos. Take any nutrition figures from the packaging in the photos, not from what I have written.

The second sentence is short and it is the whole safeguard. It also gives you a way to check the result: ask which photo a figure came from, and there should be an answer.

What does Vizard Agent actually do?

It treats the photograph as a document. When a figure needs to go on a card, the relevant part of the original image is enlarged until the print is legible, read, and the card is built from what it says — not from the brief and not from a plausible value.

In that session the sequence ran:

The figure was fixed twice, and only the second fix was reliable — the first was a correction, the second was a reading. That distinction is the whole point.

What does the result look like?

An ad whose stated numbers match the product sitting in your own photographs. The cards carry only the wording you approved, and every figure on them can be traced back by Vizard Agent to a specific image and a specific panel.

The framing ends up serving the same purpose. In that session the crop was adjusted repeatedly so the whole wrapper stayed in shot, and the montage settled on three framings of one photograph so a single package reads consistently — a cropped wrapper makes the claim unverifiable to anyone watching.

When does this not work well?

When the panel is not legible in any photograph. A dim or angled shot of small print cannot be enlarged into readable text, and inventing the value is worse than omitting it. Vizard Agent will say the figure is unreadable rather than supplying one.

Regional variation is a real trap. The same product can carry different figures in different markets, so a photo of the pack you have is authoritative for that pack and not necessarily for the one your audience will buy.

And it does not make the claim lawful. Reading a number correctly off a package is accuracy; whether you may state it in an advert is a separate question with its own rules.

How do you fix a result that came back wrong?

If a figure is wrong, ask Vizard Agent where it came from before asking for it to be changed. If the answer turns out to be your own brief rather than a photograph, that is the actual bug, and the fix is re-reading the panel rather than adjusting the number you disputed.

If the print cannot be read, send a closer photograph. One sharp shot of the panel is worth more than six of the product, and it is the cheapest fix available.

If the wrapper is cropped in a shot, say which one. Vizard Agent re-frames to keep the package whole rather than nudging the text around it, because the package being visible is what makes the card credible.

How does Vizard Agent compare to doing it yourself?

Nobody sets out to put a wrong number in an ad. It happens because the figure is typed once, early, from memory or from a supplier's summary, and after that it is just the text in the project — nothing in a normal edit ever asks where it came from.

Making the photograph the source closes that loop. The panel is already in the files at higher resolution than the video needs, so the check costs one enlargement, and it turns a remembered number into a read one.

Common questions

Does the panel need to be readable in the final video? No. It only needs to be legible in the source photograph.

Can it read a curved or angled label? Often, yes. Vizard Agent will say plainly when the angle defeats it.

What if my brief and the packaging disagree? The packaging wins, and Vizard Agent tells you there was a conflict.

Can I see which photo a figure came from? Yes. Ask Vizard Agent and it names the image.

What about the ingredients list? Same method. Vizard Agent reads it off the panel rather than paraphrasing.

Will it invent a plausible number? No. Vizard Agent reports the figure as unreadable instead.

Does this slow the job down? Barely. Vizard Agent needs one enlargement per figure.

Can it check a price too? Yes, where the price is on the packaging or on a page you supply.

What if the pack differs by country? Vizard Agent flags it as a limit — your photo is authoritative for your pack.

Should the wrapper stay uncropped? Yes where a claim is on screen. Vizard Agent frames for that.

Can it clean the background of the photos? Yes. Vizard Agent removes clutter before building the montage.

Will the cards only say what I approved? If you list the permitted wording, Vizard Agent verifies the rendered cards against it.

What if there is no panel in any shot? Supply the figure and accept it is unverified, or send Vizard Agent a photo of it.

Does it check the delivered file or the plan? The delivered file. Vizard Agent reads the rendered card, not the script.