I’ve been asked to comment on many views for planning and it’s surprising how often I’m called in to check another visualiser's work because someone in the team says, “The AVRs just don’t look right.”
I used to get pretty defensive about this remark on behalf of the architectural visualisation (or archviz) community. The renders often looked great, so the arrogant nerd in me would always want to fold his arms, pull a wry smile and tell the doubters that the work was good, and if they couldn’t see it, then maybe that said something about them. And worse, if a non-archviz (ie normal) person complained that the “perspective looks distorted”, honestly, I would suppress a scream.
But, after a long conversation with a particularly nice (read: patient) client who’d studied photography, so she knew what she was talking about, I looked again. Whether we, the ‘makers’ of these images are right or not, if the ‘users’ – the teams that rely on verified views during planning applications – think they look off, then it’s our duty to address the issue. Whether we agree or not.
As I see it, this perceived “wrongness” comes down to two issues.
1. The use, or not, of tilt-shift lenses
2. Fields of view, or more accurately angles of view.
Now, even from all the way over here where I sit at my
keyboard, at least two screens and the internet away from you, I can see your
eyes glazing over; the curled lip of a stifled yawn. But there’s an important
reason why you might want to stay with me on this. Promise.
1. Tilt-shift lenses
A tilt-shift lens lets us get the top of a building into the frame without tilting the camera. When a camera is tilted, a simple perspective of two vanishing points becomes three vanishing points and all the verticals converge to that third point. Everything goes wedge-shaped.
It’s not just that we’re used to two-point perspective from every oil painting ever, but our brains – which receive images from the eye in three-point perspective on a curved surface – correct it for us too in the process of making the world less nauseating to navigate. As a result an uncorrected three-point perspective image can look distorted on the page or screen. It doesn’t match our perception of the real world. In reinstating the order of two-point perspective, a tilt-shift lens goes some way to matching the remapping our brain undertakes.
But even a corrected perspective can look distorted. As the tilt-shift lens moves the horizon down the page, the tops of objects get increasingly wide-angle and pointy. The square corner of a skyscraper for example, can turn into a knife edge.
The reason for this is angle of view.
2. Angles of View
A while back (1995 or so) we were asked to build a ‘mini-Imax’ theatre for a property client, and render an immersive walkthrough of their proposal to run on it. I knew we’d have to render multiple panels for each frame of animation (this was before HD) and join them together to fill a screen that size, but how many panels, and more importantly what view angle did we render to fill the audience’s vision?
The total angle we can see is about 120-degrees, so if projected images span 150-degrees then our peripheral vision is filled. Once this happens, we are tricked into feeling as if we are within the frame.
We managed this with three stitched together panels for each
frame, and the effect of our first experiments were so powerful our clients had
to hold onto their chairs to stop themselves falling over when the animation moved. The
(calmed down!) version was compelling enough that the building was let before
it was built, and off the back of it Foster and Partners asked us to create an
immersive experience, this time filmed, of Chek Lap Kok airport.
So you would think that an image of 120-degrees would be optimal, and it would if we were to print it out and stand in just the right place to have it fill our vision. But this is rarely practical. We have to accept that anything less than filling our vision is a compromise of some kind, and the difference in perception is huge.
Photographers have determined that for most purposes the normal or standard lens, with a view angle of around 40-degrees to 60-degrees, represents the most natural perspective in terms of the least distortion and best separation between foreground and background.
Why this might be is up for debate. Some say it’s when the distance between observer and object equals the diagonal field of view at the object's plane. Others say it’s because a photographic print of something in this range can be held up, at arm’s length, to overlay the real scene. Some say it’s just an accepted standard based on lenses originally shipped with the most common cameras, and we’ve just become used to it.
(Interesting side note: The latter is backed up when perspectives are shown to people who’ve never seen a perspective representation of the real world; they have difficulty understanding it at all. Also, in the early days of cinema, any change of lens angle was considered confusing, so most movies where produced with one or two lenses. The range of representation, or distortion, that we can understand is definitely increasing. This might explain why people who work with perspective get frustrated with those that don’t. Though why those that don’t understand perspective seem to be in charge of all the money is still a mystery.)
Whatever the reason, 40-degrees to 60-degrees is an accepted standard to the extent that anything outside this range can be legally challenged as distorting the scene.
Here, you can see the effect of crossing a road and using a wider angle. The furthest camera position in this video uses a 45-degree lens. By standing closer to the foreground building, and using a wider lens so the smaller building stays the same height in the frame, the tower appears to shrink to insignificance.
Likewise if you ask what the view angle is, and you get a shrug and/or “It’s a 24mm lens on a Canon DSLR with a whatever up its whatever” and they don’t know what that means in terms of the view angles of human eyesight, then they haven't considered the issue. I would find that worrying.
Either way, if a view looks off, the first thing to do is always to crop it to the 40-degree mark, print it at A3 and hold it at arm’s length. Geometrically, at least, this is as close as you will get to looking (out of a little window) at the proposal.
Also our angle of vision might be 120-degrees horizontally but it’s only 80-degrees vertically. What we can take in peripherally side to side, we have to look up to take in vertically. So not only will the perspective of a tower be increasingly distorted, the further from the 40-degree marker it lies, the more looming it will appear in real life. Above 80-degrees and we’re craning our neck.
So beyond some people feeling that CGIs for planning look a little “off” why should we care? Everyone knows that estate agents have used wide-angle photography for years, to make a kitchen with a sofa in it look like a spacious studio apartment. We expect this visual white lie, and make allowances for that fact we’ll be disappointed when we come to viewing the property in real life. No biggie.
But when buildings are built, and they make a much more imposing impact than we imagined. When they loom over us, where all the verified views showed them as something over there, not right in our face, then perhaps we should care.
And if no-one cares enough to double check this, if we glaze over and dismiss it, then perhaps that suits those who might wilfully use view angles to mislead. Ask yourself how many times you’ve been surprised by how big a development seems, next to the visuals. And are you OK with this?
Perhaps, and call me paranoid if you want, whenever we ignore something because it sounds complicated and arcane, there is someone, somewhere, glad we’re not paying attention.