How to explain a technical infrastructure project on video

Last updated 2026-08-24 · 7 min read

What is the short version?

An infrastructure project is explained badly by photographs on their own and expensively by commissioned animation, which is why most of them never get a video at all. The middle path is a narrated documentary built from your own site material, with overlay cards carrying the technical labels that the pictures cannot.

1. Go to Vizard Agent and upload your photographs and site clips.
2. Say what the project demonstrates and to whom.
3. Say the register and the length.

What do you need before you start?

Site material and a clear mechanism. Vizard Agent writes the narration, sources the music and builds the graphics, so the photographs are the only irreplaceable input. Stating the mechanism plainly matters, because the overlay cards are built from it stage by stage.

• Your photographs and clips. From the site, at whatever quality you have.
• The mechanism. What the system does, step by step.
• The audience. Council, client, public, students.
• The register. Academic documentary is a real and useful instruction.
• The length. Ninety seconds covers a system properly.

What do you type into Vizard Agent?

Describe the mechanism rather than the video. Vizard Agent turns each stage into a scene and an overlay card, so a brief that walks through how the system works is doing the structural work as well as the content.

Prompt

Variants worth knowing:

• Overlay cards per stage. Clean scientific typography rather than marketing graphics.
• Chapters. For a system with distinct phases.
• A specific narrator accent. It sets the whole register.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real infrastructure explainer. The problem it caught near the end is the kind that ruins a delivery quietly: one uploaded clip carried rotation metadata that did not match its content.

1. Downloads and inspects every provided image and clip, generating annotated contact sheets with asset identifiers.
2. Reviews both contact sheets and checks the properties of the video assets.
3. Searches for academic documentary music and downloads two candidates to compare.
4. Searches for a British documentary narrator and generates the narration.
5. Extracts a precise word-timed transcript and prints the full timings.
6. Measures every audio stem in LUFS and checks the aspect ratios of the images.
7. Generates overlay cards with clean scientific typography, checks card one on frame and refines the geometry.
8. Renders a test of scene one, inspects the frame, then renders all seventeen scenes in parallel with measured pan and zoom.
9. Checks the rotation metadata of every video asset, finds one wrong, builds a split-blur portrait frame for it rather than cropping, re-renders the affected scenes and reviews two verification sheets of the master.

Step 9 is the detail worth knowing. A phone clip whose rotation flag disagrees with its content will appear sideways or letterboxed, and the fix is not a crop but a frame built around the true orientation.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1920x1080, H.264, 30fps, 95.5 seconds, AAC audio. Widescreen, ninety-five seconds, seventeen scenes built from site photographs and clips with pan and zoom, narrated, with scientific overlay cards, captions and crossfades between chapters.

Ninety-five seconds across seventeen scenes gives each stage around five seconds. That is enough for a card and a line of narration, which is the minimum a technical stage needs before it becomes a blur of images.

When does this not work well?

An explainer built from site material is bound entirely by what somebody happened to photograph, and technical accuracy is not something any edit can supply on its own. Vizard Agent builds the structure, the pacing and the labels, and the claims underneath them stay yours.

• The mechanism has to be described correctly. Vizard Agent narrates what you tell it.
• Site photographs vary in quality. Vizard Agent animates them; it cannot improve them.
• Nothing here is a simulation. Pan and zoom on a photograph is not a model of behaviour.
• Rotation metadata lies. Vizard Agent checks it, and unusual capture devices remain a risk.
• Ninety seconds is an overview. A full technical case needs a longer format.

How do you fix a result that came back wrong?

Name the scene or the card. Vizard Agent keeps the annotated asset sheets, the narration timings, every overlay card and the verification grids, so a corrected label or a replaced image is a re-render rather than a rebuild.

• "That stage is described wrong." Corrected and the narration re-recorded.
• "The card overlaps the detail." Geometry refined and checked on frame, as here.
• "That clip is sideways." Rotation checked and a proper frame built for it.

How does Vizard Agent compare to doing it yourself?

By hand this means building seventeen slides in a presentation tool, recording a narration over them, and accepting that the technical labels sit wherever the template puts them. The sideways clip is the failure that survives review, because it looks correct in a file browser and wrong in the export.

Common questions

Does this work for other technical subjects? Yes. A manufacturing process, a water treatment plant or a construction method all explain the same way: Vizard Agent gives each stage of the mechanism its own scene and its own labelled card, built from whatever site material you actually have.

Do I need drawings or models? No. Vizard Agent works from your site photographs and clips, animates them with measured pan and zoom, and builds the technical labels as overlay cards on top.

Will Vizard Agent explain the engineering correctly? It narrates what you describe. The technical accuracy is yours to supply and check.

How long should an explainer be? Around ninety seconds for a system with several distinct stages, which gives each one roughly five seconds on screen.

Can Vizard Agent use a specific narrator accent? Yes, and it searches for one that matches the register you name.

Why not just crop a sideways clip? Because cropping throws away picture and often leaves the subject half out of frame. Building a split-blur frame around the true orientation keeps the whole image and looks deliberate rather than salvaged.

Does Vizard Agent add captions? Yes, in a documentary style, positioned after checking them on frame.

Can Vizard Agent handle a mix of photos and video? Yes. It probes both, annotates them with identifiers and builds scenes from either.

Does Vizard Agent check the finished video? Yes. It reviews two verification sheets covering every scene of the master before delivery.