The hidden cost of photorealism

Time to ponder on what nobody cares about, as usual 😊

I recently finished a newsletter article and figured, why not share this one here as I think it is actually quite relevant to the evolution of archviz.

Last month a report was released by cgarchitect that showed something that didn't surprise me, but that still made me ponder the subject for a bit longer. 

While the tone in the report makes it sound like a cool thing. I think there is a genuinely concerning obsession with photorealism in the industry. 

This of course is not new. Since rendering emerged, people have been striving to make them look more and more photorealistic, more convincing, more lifelike. Using render engines used to be a very technical tasks, and one would spend more time finetuning subdivs than actual component of the image. This is actually this technicality that made this service being outsourced from the get go, as clearly architects and property developers would never have the bandwidth to use such tools to a decent level. 

Fastforward a few decades, rendering engines are now much more user friendly, even seasoned professionals rarely go under the hood to fix anything because the default settings work 99% of the time and the hardware has evolved so much that a rendering that would take a whole night to render will now render in half an hour. 

This jump in productivity came also with a leap in realism due to better engines that "fake" the light more accurately even with biased render engine. Ready-to-use libraries getting bigger and bigger also helped in achieving realism faster with better materials, and high-polys assets.

Funnily enough, our perception of the "standards" just evolve with what is available. Renderings from 20 years ago actually looked as realistic back then, as new renderings look realistic now. Of course, if we look at renderings from 20 years ago, they look like hot garbage. It's just a question of perspective, and comparison. 

A perfect example of this is the field of video games. I remember being blown away the first time I played Half Life. 

I was thinking this was "as good as it gets". And back then, it indeed was. Then came Half Life 2 that, again, looked "lifelike".

Of course, while this still "holds up" nowadays, which is impressive, we know we can get much better graphics today. 

Same with visualization. Some renderings done a while ago (I'm thinking of some cool Bauhaus stuff by Bertrand Benoit) were already pretty impressive, but we know we can do better today in terms of photorealism. 

Interestingly, architectural visualization is taking the exact same route as video games. 

There are still some cool games coming out every now and then, but overall most AAA games are advertised as technical demos, because the gameplay is absolute garbage. Problem is, the reason we play videogames is mostly for the gameplay. A game that looks realistic but is boring to play will not keep the players hooked. 

Historically you have tech demos that were also fun to play (Far Cry and Crysis for instance). They also had an incredible art direction. The scenario was not mind blowing, but it kept you going. But this is less and less the case. 

Same goes with the landscape of architectural visualization. More than 50% of practitioners seem to think that photorealism (which is different than realism, but that's another topic) is extremely important. 

To me, if something is "extremely important", it basically will take most of my production time, and everything else becomes secondary and will not have any sizeable chunk of my schedule. 

There's at least two issues with this approach. 

The main one is to consider photorealism as an end rather than a mean. 

Written like that it seems rather simple and stupid. But this really sounded like an epiphany when I finally put it into word. This obsession with realism means that it has become the end goal of an image. An image is good if and only if it is as photorealistic as it is possible. The quality of an image is entirely based on its photorealism. The lens through which people will look at an image, is only to nitpick on potential "unphotorealistic" elements. 

Problem is, if I take my phone and take any shitty snapshot, it is photorealistic. But it's a shit image. 

And the same happened with archviz. The obsession with photorealism has lead to increasingly photorealistic images, that also became dramatically cliché and boring. Because once you're happy with the realism of your image, and only judge your image by this metric, then you think your image is good, and you call it a day. 

Just like we used to be happy with the realism of Half Life almost 30 years ago, we tend to be happy with the ambient realism of images of today. Only the eye of a professional can now see the remaining gaps between what we have now and a truly photorealistic image. 

But the thing is, we don't make images for professionals of the field. We make images for professionals of fields that have absolutely nothing to do with this issue. What this means is that the expectation of what "photorealistic" is from a client perspective, has been reached years ago. Yes there are a small subset of clients that are extremely demanding on the realism of renderings, but those are far from the majority. And I would actually love to do some A/B testing with those demanding clients, because I sometimes think that it's not that they are demanding for "more realism", but simply that they like sending reviews and asking for change. 

We all know about those stories where a client asks for tiny changes, and the artist sends them the exact same image an hour later saying they've implemented the changes, and suddenly the client is happy "Aha! You see, now it's perfect!". 

All the more importantly, the actual end-viewer (the true "client", or buyer) is even less aware of those potential gaps between the image and what it could look like if truly photorealistic. The vast majority of people think AI images are already lifelike even though professionals can tell them from a few miles away. 

Professionals are nitpicking on what doesn't make any difference. While this in itself is not a problem - people can dedicate their time to whatever they want - the problem is that they do think that it makes a difference, whereas we clearly know it makes absolutely none. 

Funnily enough, a layperson will be able to tell if an image is boring. In a split second even.

The primary vision will not pickup on potential "unphotorealistic elements". The layperson couldn't care less about wrong bump map multipliers, some fireflies or a noisy shadow.

But it can surely tell if what is happening is pleasing, if the colors are making them feel something, if their eyes are navigating in a dynamic or restful way in the canvas, if the approach to the topic is engaging or if they've already seen this a million times.

In a nutshell, the end viewer can tell if your image is worth their time. And photorealism is much lower in their priorities than artists would like to think. 

I remember when I started working on my course that the core driver for me was to try to refocus the attention of artists on what matters, to make them ponder topics that most thought irrelevant, and to show how much impact it has on the process of crafting images.

4 years later and I see that the problem at hand is still deeply ingrained in most artists' brains, and that its impact on the quality of the images being produced is more and more noticeable.
 
Photorealism is a mere subset of the overall quality of an image. Images themselves are also a subset of the actual service offered by archviz artists.

Photorealism is getting easier and easier to achieve, it is time to focus back on what matters, on what makes a difference. Photorealism is simply not enough anymore. Time to shift focus.