Why Every Flythrough Looks the Same

I was scrolling LinkedIn this week, when I ought to have been doing something more useful, and came across four or five architectural animations in a row. Different studios, different schemes, different parts of the country.

All competent. All beautifully lit. And I couldn't have told you which was which an hour later if you'd paid me.

Archviz Slop

There's a phrase doing the rounds at the moment: AI slop. The flood of content that's technically fine, instantly recognisable as machine-made and entirely without a point of view.

I'm tempted to say our industry has its own version, though that probably isn't fair on the people making them. Most of these films are good work by people who know exactly what they're doing. It's that when you watch several in a row, the formula becomes rather obvious.

Which is what makes it awkward to talk about. The lighting is right, the trees are right, the people crossing the plaza are right. The problem is that you could swap the building for a different building on a different site and the film would be more or less unchanged. Competence without an argument.

We're reworking one at the moment that somebody else produced. It's a perfectly good film, and it shows the whole development: every elevation, the yard, the entrance, the parking, the lot. What the client wants to talk about now is the landscape, the planting and the walks and what it's like to wander around the place on a Sunday. That gets about four seconds somewhere in the middle, from four hundred feet up.

The film answered every question except the one being asked. It showed everything, and that was the problem.

Where the First Shot Came From

So why does it keep happening?

Architectural animation borrowed a good deal from cinema, and the establishing shot most of all. Open wide, give the audience their bearings, move slowly in towards the building.

In a meeting that's exactly right. Somebody has introduced the scheme, everyone knows why they're watching, and there's usually an explanation either side of the film. The wide opening orients people who came to be oriented.

Nobody in a feed came to be oriented. They weren't introduced to anything, they weren't looking for your development, and more often than not the sound is off. Spend the first few seconds approaching the site from above the clouds and they'll never reach the thing the film was made to show them.

Which is the part I think gets missed. A film doesn't arrive in a vacuum: in a meeting somebody introduced it, on a website the viewer went looking for it, in a feed neither is true. All three tend to get the same opening shot.

One Film, Four Audiences

Here's the part I'd defend rather than criticise.

Most animations have to work for planning, for investors, for the agents and for social media. Four audiences, four quite different needs, one budget. What comes out is a compromise between four briefs, and since most schemes are compromising between the same four, the films land in roughly the same place. How many of the four got what they actually needed?

That's a scope decision rather than a creative failure, and it isn't necessarily the wrong one. If there's money for one film, versatility may well be worth more than originality. It's worth knowing what's been traded away: a film that works reasonably well everywhere tends not to work especially well anywhere.

The alternative isn't four films. It's deciding early that one production can yield a long version for presentations, a shorter one for the agents, and something that gets to the point for social. Re-editing footage you already have is a far smaller job than creating it. But it has to be planned for rather than hoped for afterwards.

And it still surprises me how often a brief names the technique. "We need a flythrough" describes a camera movement. It says nothing about what anybody should understand or feel, so what comes back is flythrough grammar, delivered exactly as asked for.

The Defaults Are Somebody Else's

I'd assumed AI would break the convention. So far it mostly reinforces it.

These tools can be steered, with references and keyframes and a fair amount of control. Left to their defaults, though, they return what they were trained on, which is the average of everything already made. There must be ten thousand flythroughs in that training data, nearly all opening on sky. Ask plainly and you'll get the ten thousand and first, rendered immaculately.

There's a small and very British proof of this. Ask for a road through a UK development and more often than not the traffic comes back driving on the right. Not every time, but often enough that we check for it. The tool has learned what a road looks like from everywhere except here, and it has no idea that this is the one detail a British viewer will catch in half a second.

The average of everything isn't neutral. It's just somebody else's normal.

What the tools have genuinely changed is how cheaply you can ask a question. Testing an odd camera move costs a fraction of what committing to one used to. What they haven't done, in our experience, is make a drawing-accurate film cheaper to produce, and I'd be careful about assuming the next one will.

Breaking the convention is a different matter, because the deciding costs nothing. The shot itself might, and a close-up asks a good deal more of the modelling than a distant aerial does. But choosing to open somewhere other than the sky is free, and it has to happen before anyone starts animating.

When the Formula Is Right

I should be fair to the formula, because it exists for a reason.

A film explaining a scheme to a planning committee has a different responsibility. The context has to be clear, the material has to be credible, and nothing should get in the way of what members are being asked to assess. An engaging opening isn't wrong in that room, but it has to serve that job rather than distract from it.

So the convention isn't wrong. It was built for one room, and now it gets used in all of them.

The One Thing

The most useful film briefs I've worked on have one thing in common. At some point, somebody has been able to say what the viewer should still be carrying around when it finishes. Sometimes that arrives with the client, and sometimes we find it together, which is half of what the early conversations are for.

Not the format. Not the running time. The one thing.

On the landscape animation we're reworking it's easy enough to say out loud: this should feel like somewhere you'd enjoy walking around. Once that's the starting point, the choices about camera and pace and order more or less make themselves, and the film looks nothing like the one it's replacing.

Everything else is a camera move.

Until next week.

Rich

PS: One day I'd like to make an animation that opens on a door handle (or a bumble bee!). No sky, no context, no slow push in. Just a handle/bee, and then we'll see what happens. If anyone fancies being the one to try it, do get in touch.