Originally posted by: mwmorph
Originally posted by: Kakkoii
Originally posted by: mwmorph
Originally posted by: Kakkoii
Originally posted by: mwmorph
Originally posted by: Qbah
I'd say to have something like
this rendered in real-time at an acceptable framerate? Link from PC Gaming forums
I find it funny that people keep saying, graphical quality is stagnant. When you get to the point of Crysis, there really is little place to go and the wall of diminishing returns comes in. Vegetation and lighting already looks so lifelike. The nanosuit looks great, hell almost all character models look realistic (ignoring human faces, which man has never been able to artifically duplicate in any medium convincingly). The only place in improvements might come into play that really tend to break the immersion are human face textures which will almost be impossible.
If the 5870 or GT300 allow gamers to run high rez, high detail smoothly in games that look like Crysis, I really thing that beyond that, purely graphical wise, there's little reason to upgrade any further in the foreseeable future.
Sigh... You are sooo wrong.
You may think things look about as good as their going to get, but truly there is a lot more room for improvement. Things like particle physics & physics in general are horrible in most games currently. It takes a lot of processing power to do properly. The more processing power we have, the closer to replicating reality we come. There's no wall of diminishing returns. Just a wall of processing power.
And I don't know where you've been the past 8 years, but human faces have been replicated perfectly. Nvidia even has a tech demo for 2009 that is quite convincing, if you showed it to someone who had never seen it before, most would think it's a real person.
Nvidia 88000 Tech Demo
http://www.youtube.com/watch?v=LIGWAYS5uRw&fmt=18
PS3 Alfred Molina demo
http://www.youtube.com/watch?v=W9QbYWEd7AE&fmt=18
Image Metrics CG Facial Animation
http://www.youtube.com/watch?v=UYgLFt5wfP4&fmt=22
Pendulum Studios Alter Ego division
http://www.youtube.com/watch?v...fmt=18&feature=related
Liam Kemp's facial animation test
http://www.liamkemp.com/photoshoot.htm
And that's animated stuff. 3D modelers can make perfectly real looking models.
Ultra realistic korean singer model
http://maxedwin.cgsociety.org/gallery/399499/
http://features.cgsociety.org/..._1110764412_medium.jpg
It's processing power that is holding us back. When we have the power to realistically light a scene, things will look so much more real. Whether it's by ray-tracing or something similar. Right now though it takes a bit to much processing power. But ray-tracing on a GPU is making great progress and performance already.
Sigh... someone needs to learn how to read. That's why I mentioned diminishing returns right at the beginning. You can put 100x more processing power for a moderately better looking image, but at that point, the cost and time required to do so make it uneconomically feasible anyways. Yeah, chain 5 million 4870s together. You'll get your real life graphical quality, but you'll never find a dev house that can do it on an economical basis to create a game for it. Art costs money and time. AAA titles today can cost up to $100million to develop already, who is going to put in the exponentially higher cost and time to make something look like that?
I won't go even into detail about why I need to draw a line between animated and static modeling. I think it should be patently obvious the difference and challenges the former that the latter can so easily avoid from a psychological perspective.
It' not power holding us back. A couple more generations an it'll be the practicality of the real world holding us back.
Uh no, with more processing power it takes less time. Because the hardware is then capable of running software that does the hard work for the designers. Such as making things move realistically, and making the lighting look great. Instead of spending tons of time tweaking the different lighting effects and employing tons of tricks to try and get it as close to photo realistic as possible. More processing power allows for much more automation of effects instead of creating them beforehand and having them replay in the game.
Let's say we had the power for large scale real time ray tracing in games more complicated that Crysis. And we had an engine that combines ray-tracing, physics and destructible environments. In said engine, all developers would have to do is assign a pre-defined/tweaked material and molecular structure and to an object. The engine then does the correct lighting, physics and breaking correctly for the developers based on the materials chosen that make up your model.
Sure, if a developer wanted to create an uber realistic game with all that processing power from freaking scratch, then yeah it's going to cost too much for them. That's the point of freaking MIDDLEWARE!
Ray tracing... There is nothing rasterization engines cannot do that raytracing can except for 100% realistic self reflections and there's many things ray tracing can't do that rasterization can like real AA. Ray tracing is not some panacea, it's not something you jut say, ok, everything is fixed now. Raytracing creates just as many problems as rasterization does.
Reflections with multiple bounces is the prime reason why Ray-tracing is very good. It's the touch that makes a scene feel so realistic. Because light is being refracted/reflected in all the right ways. Ray-tracing replicates how light act's in real life basically. It simplifies the work needed to make something look real in a scene. If you've ever used Ray-tracing in Maya, you'd know that. And you'd also know that you can Anti-alias with it. I do it all the time.
Also, ray-tracing does transparency way better than rasterization ever can.
And ray-tracing probably isn't the panacea. I'm not saying it's 100% perfect. Something better may come along in the future. But for now it's the best at creating realistic scenes. And with Chaos Groups V-Ray for GPU, it's getting ever closer to real time.
Originally posted by: mwmorph
You're also back to the exact wrong points, what you mentioned is all engine wise, programming. What I was talking about is art. Who is going to do the million+ poly models? You cant tesselate complex shapes perfectly without work. Who is going to work on the extreme resolution textures an different maps? You still have to create the massively more detailed objects, you still have to texture them, then map them, then adjust every detail such a surface diffusion, reflection, refraction, subsurface scattering, etc to make it look 100% realistic?
Ever heard of Zbrush? No? You should try it. lol.
And texturing wise isn't really much of a problem. But it is a problem that will be solved through software. Further down I'll go more into it.
"surface diffusion, reflection, refraction, subsurface scattering, etc"
Those things don't exactly take days to adjust.
Originally posted by: mwmorph
Have you ever set up a 3D scene in Maya or 3dsmax? I have. Creating complex shapes take time. texturing takes massive time, an as you reach the boundary of lifelike, each 1% improvement takes more an more time. Now to achieve this holy grail of absolute realism you keep trying to push, I would have to do that for every object in the game, creating unique models and textures for every face. I would nee to subtly re-texture every door, wall, crate, down to something like 15k x 15k pixel texture maps since no two things in real life ever look the same. There is no point, just like programming, all you can say at some point is from a practicality point of view, it's good enough and you cant afford to and it wouldn't be humanly possible to do more on a large scale like a AAA title game.
So have I in Maya. But I don't create complex organic shapes in Maya, I create them in ZBrush, where it takes WAY less time to do. And allows your creativity to flow.
You don't have to do that to every single thing in the game. That's why I was talking about middleware. You have software that creates these random variations, imperfections for you based on a bit of input first of course.
Watch this video and you'll get an idea of what I mean. It's near the end of the video:
http://www.zbrushcentral.com/z...howthread.php?t=074539
The noise feature they will be implementing is sort of what I mean. The same as real life, random imperfections.
The same would have to be done for basic types of materials that are constantly found throughout a game world. Such as wood beams for example. You would have an algorithm for creating random variations in the grain of the wood. Obviously constrained by certain variables to keep it looking like wood of course though.
But I would go even further than this. Use the ray-tracer or w/e to create the colors of the object by assigning some sort of randomized molecular structure code map on the objects, that the ray tracer would then know how much refraction, reflection, diffusion, etc.. to do based on it's real life counter part. Dunno though, this is probably a wacky idea lol.
Basically what I'm saying is that these problems your talking about can be solved through software. So that the artist's are still free to create, but the long remedial stuff is easily created through randomization based on a bit of user input and also defined presets of tons of different materials that are used most in real life.