How to animate a scene set at a real building
Give Vizard Agent a building's name and address and say what should happen there. Vizard Agent looks the place up, pulls exterior photographs from its website and from map listings, builds a reference sheet from what it finds, and generates the scene so the building in it resembles the real one.
What is the short version?
Generated video will happily produce a church, a school or a shop. Producing your church, school or shop is a different problem: it means giving the generation something real to work from, and for most buildings those pictures already exist online — on the organisation's own site and on map listings.
- Go to Vizard Agent and name the building and its address.
- Say who is in the scene and what they do.
- Say the length and the shape.
What do you need before you start?
The address and the action. Vizard Agent finds the photographs itself from the organisation's site and from public map listings, and what it needs from you is precision about which building you actually mean — a name without a street and a town is a different building somewhere else entirely.
- The full address. Name, street, town. This is what makes the lookup unambiguous.
- Who is in the scene. A family, a couple, a class. Describe them briefly.
- What they do. Walking in, gathering outside, arriving in the morning.
- The time of day. It changes the light and therefore the whole feel.
- Your own photographs, if you have them. They beat anything published online.
What do you type into Vizard Agent?
Name the place precisely and the action plainly. Vizard Agent does the research itself, so a single sentence with a real address in it is a far more useful brief than a long description of a building it would otherwise have to invent.
Prompt
Variants worth knowing:
- Upload your own photos. Vizard Agent uses yours before anything it finds online.
- A different time of day. Say "morning light" or "evening"; it changes the mood entirely.
- Wider or closer. A wide establishing shot and a close arrival are different videos.
What does Vizard Agent actually do?
Here is the order Vizard Agent worked in on a real location scene. The first half of the run is research: it does not generate a generic building, it goes and finds what this one actually looks like.
- Loads the film playbook and lists the available assets.
- Looks up the real building's appearance.
- Checks the cinematic and image generation options.
- Fetches the exterior from the organisation's own website.
- Searches for building photographs and street-level imagery.
- Screenshots the map listing and extracts the images from it, then looks at the screenshot.
- Downloads the exterior photographs and reviews them, including a close-up.
- Creates a combined family and location reference sheet and reviews it.
- Generates the arrival scene, tiles its frames for review, analyses the motion and audio, and mixes music under the final render.
Steps 4 to 7 are what make this specific rather than generic. Vizard Agent assembled a picture of the real building from several public sources before generating anything, which is why the result resembles that place rather than a plausible one.
What does the result look like?
From the run this page is written from, probed on the delivered file: 1280x720, H.264, 24fps, 10.1 seconds, AAC audio. Wide, ten seconds, a generated arrival scene built against reference photographs of the actual building, with music mixed underneath.
Vizard Agent listened back to the final mix and reviewed the frames as a contact sheet before delivering, rather than uploading the first render.
Ten seconds is one shot: an approach and an arrival. It is enough to establish a place and put people in it, which is what this kind of video is usually for — an announcement, a welcome, an invitation.
How long it takes was not measured separately here. Comparable work runs a median of 28 to 38 minutes end to end, and the upper quartile is several times the median. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.
When does this not work well?
Depicting a real place is a claim about that place. Vizard Agent builds a resemblance out of public photographs and checks it against what it found, and a resemblance is still not the building — and the building belongs to somebody who has a say in how it appears in a video.
- It resembles the place; it is not the place. Details will be wrong. Do not present it as footage.
- Use it only for your own organisation. Animating somebody else's building into a video is their decision.
- Published photographs may be dated. If the building was repainted or extended, the generation follows the old pictures.
- Poorly photographed buildings generate poorly. If the only image online is a low-resolution thumbnail, the resemblance suffers.
- Generated people are generic. They are a family, not your congregation. For real people, film them.
How do you fix a result that came back wrong?
Say what is wrong with the building or the action. Vizard Agent keeps the downloaded photographs and the reference sheet, so regenerating with a different angle or a different time of day works from the research it already did rather than starting the lookup again.
- "The entrance is on the other side." Regenerated against a different reference photo.
- "Make it morning." A new generation from the same sheet.
- "Show the sign more clearly." A closer framing on the same building.
How does Vizard Agent compare to doing it yourself?
By hand this is a photographer on site for an afternoon, or a generation tool that hands you a perfectly nice building that is not yours. There has never been a middle option that costs nothing and still resembles the actual place you are trying to show.
| By hand | Vizard Agent | |
|---|---|---|
| Getting the building right | Photograph it yourself | Researched from website and maps |
| The people | Cast and film them | Generated to your description |
| Consistency | Not an issue | One reference sheet for both |
| A second version | Another shoot | Regenerated from the same research |
Common questions
Does it need photographs from me? No, if the building is photographed online. Send your own if you have them — they are always better than what is published.
How close is the resemblance? Close enough to be recognisable, not exact. Vizard Agent builds from reference photographs rather than a model of the building.
Can it use a specific time of day? Yes. Say morning, midday or evening and the generation is built to it.
Can it show people I know? No. The people are generated. For real people, film them and Vizard Agent will cut that footage instead.
How long can the scene be? Around ten seconds. Longer generations drift further from the reference photographs, and the building starts to stop being yours.
Can I use it on the website? Yes, as long as it is your own building and you do not present it as photography of the place.
Can I get a vertical version? Yes. Vizard Agent generates it against the same reference sheet rather than cropping the wide one.
Does it add music? Yes. Vizard Agent searches a licensable library and mixes a track under the scene, then listens back to the result.
Can it show the inside as well? Only if photographs of the interior are published or you upload them. Vizard Agent works from what it can actually see.