True Photorealistic Graphics - when to expect? Maybe never?

I think realism should be strived for simply because it's a difficult goal. Once each and every game has the option of having that realism, then it'll simply be a design choice (rather than technical limitation) as to whether or not that should be the look of the game.

People like to point at Sin City and say how much cooler that is than your regular 'real'-looking movie. People say how reality is boring and uninteresting. But if every single movie ever made were simply hand-drawn animation or some CG-stylized look, then you'd better believe that people would be craving for a movie that had real looking people, environments, etc.

 

I think much of the low hanging fruits have been picked already.

A pure opaque texture is able to emulate specular effects and parallax occlusion mapping quite well in the hands of a good artist. Doing it 'for real' instead of drawing it into the texture takes a huge amount of brute force that won't compare well with the improvement in looks.

Increasing the resolution by 4x does not look 4x better(not to me at least).

A 10x increase in the number of polygons of a well-made model does not even look close to twice as good.

Real-time soft shadowing, unified lighting etc. does not look so much better than just regular light maps compared to it's cost.

We have the performance, so of course well use it; and I'm not saying it should be any other way. But I think brute force is starting to get somewhat obsolete.

I think we are already at the point where time spent on art work and other aspects like clipping, character animation and AI behaviour matters 10 times more to graphics than any brute force increase like texture resolution, screen resolution, AA samples, shader instructions per pixel, textures per pixel etc.

 

I believe that by 2015 most games will have an equal to FF:The Movie graphics level.
My guess is that even Elder Scrolls V (due out 2010) will be almost there in graphics quality with little things more to desire - graphically, that is. If we are talking about physics and AI now (which we aren't since the thread starter pointed out only photorealistic graphics), then the time-plan goes well beyond 2020 to catch absolute believability.

My .02$

 

Lighting is still the limiting realism factor in prerendered movie screens.

The goal of course is to have one 'physical' algorithm that does it all for you (eg a physics simulator of optical wave mechanics or energy transport, often called the rendering equation in full generality). Unfortunately the complexity of doing it formally for all points in a scene is still far beyond what a super computer can do working for days.

So again you resort to algorithms that approximate this 'solution' eg radiosity, monte carlo approaches like raytracing, etc etc

 

Hmm, didn't got that.....

Where can I find more about this approach?

 

Google is your friend =) I found the original paper here

http://www.cs.princeton.edu/courses/archive/fall02/cs526/papers/kajiya86.pdf

along with the wiki page

http://en.wikipedia.org/wiki/Rendering_equation

 

looks like doom3 with higher res and more polys. which I welcome, and comforts me in my little idea that doom 3 is the new quake 1

first screenshot looks a bit more cartoonish (not necessarily a bad thing). shiny nose is the biggest problem. and I'd reasonably guess that second shot would look like the first one if similar lighting was applyed (such as, switch on the light, you computer nerd!)

that Tom Cruise scientist lacks anti-aliasing. Borg grey haired guy seems almost not aliased but it's because he's more downsized. someone said years ago that anti-aliasing would become like 32bit rendering (bandwith hungry but always used). Please, at least use 2x !
why should we watch jaggies on a dozens million dollars game running on a 1.21 jiggowatt console attached to a giant HD screen.

 

He's on a TFT screen within the game.

Snake looks fake, in no small part through his teeth, which aren't convincing. For the illusion, they should be made less perfect, and a bit yellow because he smokes every now and then. There are plenty tell-tale signs that he's not real, Otacon looks much better in that respect. The guys from Crisis don't show teeth, which really helps.

Which brings me to another point. At some point AA isn't going to be realistic. Say I record something from real life with a digital camera. It's going to register pixel perfect what it sees. If there is a sharp line in the camera's view, then you will, if you look / magnify close enough, see jaggies. Jaggies should disappear from view through higher resolutions, more pixels. In the meantime, using AA is a stopgap, and in my opinion, the better resolutions get, the worse AA affects sharpness and color. Of course, AA gets smarter and can create the illusion of sharpness also, if carefully applied only to edges, and not the full screen.

 

My bad

Thanks for the sources though!!!

***sits down reading***

 

Oh well

I believe this "real AI" you are talking about, will be the single most challenging aim game designers will face in the near future. We are nowhere near it, and probably we will not be in this decade.

If I had to speculate for the sake of it, then I would say that around 2020-2025 we will be near the aim. The characters will have a very human-like behaviour, they will be able to talk about many subjects, maybe subjects out of the game's strict storyline. They will share huge databases of knowledge making them able to learn as the time passes.

 

and then they'll kill us all

 

I don't think they'll have the means to do that, unless you consider the arms and weapons in Battlefield 7 enough for this job

 

I think dealing with the "content explosion" will be a HUGE challenge. I think where quickly comming up on a decade to produce a game with a descent amount of content.

I think something like a shared library of common objects in a standardized format including all nescessary information like physics hulls, textures, shaders, animations and allowing for scalable LoD might be a solution. There's literally tens of thousands of innocous objects that contribute greatly to atmosphere without ever being the center of attention. Paper clips, pens, books, torches, boxes, crates, food items, plants, trees,lockboxes, bricks, rocks, clothes, random generic people etc.

To convince anyone to share some of their effort would require an open source style licence agreement for content that's not too draconian. If it ment making your main characters and main items in the game public and free for everyone to use no one would do it. It would have to be limited to inconsequential things and still would have to manage to get enough new content from developers to be self-sustaining. You couldn't have developers who don't produce any art and just mooch of the shared library of items to create their entire world. And the format would likewise have to be very scalable.

By preferance, no one would like to do this. But I think time constraints will eventually become less preferable than allowing others to use some of your art.

 

I think that each time we try a new game that has got new features we can just tell "ok is good enough for this generation of games" but we cannot see major improvement like we played with Quake1 or Duke Nukem 3D. Maybe years ago it was easy to get shocked also with minor difference from one game to another. Nowdays it's different: with so many polygons, effects, AA, HDR, bloom, PS and "now" physic acceleration maybe we cannot see improvement because there are simply too many detail wich we are accustom of but the realism remain the same of many years ago.

 

I've always wondered what an "external viewer" would think of the human race, seeing how we have companies spending gazillions of dollars trying to reach that goal, and millions of people spending just as much to buy hardware that can't do it yet, with the expectations that "one day" we will get on a TV or PC monitor what, ultimately, we have everywhere around us every day of our lives: Reality.

Philosophically it's absolutely amazing to analyse. It's a bit like trying to fit a seafood starter, pizza, large drink and dessert all in a 1cm cube and eat it like that, instead of just having the dinner normally...

 

I think yourself know the answer to that, but i'll share my thoughts on that.

I think that if an "external viewer" ever found out about us Humans it would come to the same conclusion we do, being that we humans (from casual gamers to all related gaming industry companies) are going after photorealistic graphics, simply because reality doesn't let you race a beatifull machine and in case you loose control of it get out of it un-hurt, or use Guns and blow up "people" as if you were some hero or villan, etc.... unlike the Reality we exist witch brings heavy/fatal consequences wich would end our unique opportunity of interact, explore think and other aspects of what real life involves.

But why photorealistic you(or the "external viewer") may ask, im sure u know the answer, we humans get bored with life easly, thats why we search for entertainment and when that entertainment is getting boring to, we search ways to improve it and keep things challengeable and some what new to avoid monotenous repetition.
Improving our visual experiences is the most desirable/appealing way to excite/stimulate/entertain ourselfs, since vision is our primary sense of awareness.
Basicly that's it i think.

 

I think drugs work just as good, more quickly and most importantly, more cheaply!
kidding...

I know what you're saying, and i agree of course. It's just amazing that we should try to fit reality in a little box. Or am i the only one to find this mind-bending in its implications?

When we do get to that level, what will happen to our minds? When what we see on screen is indistinguisheable from reality, when we shoot someone in the game and everything tells our brain that the scene looks and sounds real, added to the behaviour of the murdered character too, which i'm sure will become more and more realistic and eventually will seem like the same behaviour any human would have when he's about to be killed... If someone "enjoys" that, then i'm really worried... It basically means that someone enjoys the act of killing someone and has the guts to see and hear all about it, with the difference that it's not actually real... I wouldn't want to think how this will affect someone with a "weak" mind. Some people already have trouble distinguishing what's real from what's not, i can't imagine how bad things will get.

 

I think on the one hand thats a smart idea, but in another way it could e a problem. I think that would do for game content what UE3 has done for rendering style (very samey).

My POV is that if you have a database with all these objects and preset stats and textures associated, then you may get stuck with databases that dont update as rapid as the tech that can run it?

I guess somethings got to change though, and you are right, we need to start thinking with different views on the long term goal.

 

L-B,

I think there is a chance you were actually on drugs when you wrote that Could you spare some for me? LOL !