Hyper Photo Realism

I watched this video about photorealism in video games[a], and while I think the author makes a good point in that we're still so far away from actually achieving it in a way that is practically applicable to interactive entertainment that there is no point in building something on the assumption that what we have will be "good enough", it misses the both the point of why photorealism is useless and the point of why it still makes sense to strive for it.

Photorealism is basically making something artificial - typically something computer-rendered - look real. This takes considerable skill and computing power, and has so far only been done in limited ways. In the scope of anything aspiring to be art, however, it is very much a red herring.

Even in photography, an art where it is difficult not to be photo-realistic, we don't try to depict the world as it is - even when we do. When looking at a scene we don't observe every part of the scene in the same way - we focus on some elements and ignore others. When we look at a sunset, the bright red disc of the sun is certainly perceived as being larger than the bright spot overhead at noon. When going from that scene to a finished JPEG that is uploaded on this blog, it isn't sufficient to numerically preserve the colors, the highlights and the shadows - the relative prominence and relations of the elements in the scene must also be preserved and communicated to the viewer if the end product is to have the desired impact.

The photo won't say "sunset" unless it is made to say that.

The same goes for interactive entertainment. An alien world must be made to look alien. Otherwise, no matter how much it looks like the real Mars, it will still be perceived as a sand heap seen through orange-brown goggles. A desert must be made to look like a desert, and not just like a place with some sand happening to be lying around.

Image taken by NASA's Curiosity Mars rover just north of the "Kimberley" waypoint shows beds of sandstone. (Source: NASA/JPL-Caltech/MSSS[b])

Photorealism deals with reality, but art is at most curated reality. That is why it is largely useless in the creative arts - ultimately, nobody wants the result.

Photorealistic rendering techniques, however, will give us the level of control of the image that we require to make the image say what we want it to say without making it cartoonish. Having a system that is sophisticated enough to make something look real is a great jumping-off point for making art, which, being a kind of curated reality, can be considered hyper-real.

This is where I see photorealism ending up in the context of art. We will continue to research photorealistic rendering, but only because the way to the hyper real goes through the real. It's only interesting as a point of reference that we can start our exploration at.

Finally, an example of photorealism merely being a starting point is the demo Photon by Fairlight[c]:

The coder, smash, managed to do wonderfully photorealistic rendering by a technique called "photon mapping" where a multitude of light rays are traced through the scene, a very difficult thing to do in real-time. Still, he saw the need to add a visualization of the photons bouncing around, and the reception of the demo was... well... Smash admitted that unless one saw and could appreciate the technical skill that went into it, there wasn't much there.​[1] Photon found itself in second place. The winner? Coronoid by Still, an abstract explosion of hypercharged neon polygons and otherworldly landscapes.

For reference, a still from the ending climax of the demo. See the full demo here.

Footnotes