AncientSpaceAeon Posted May 24, 2011 Share Posted May 24, 2011 (edited) (The original title was, "Is it just me, or that no matter how they tried, the developers can't make it look "real enough." Too much characters, perhaps.) The graphics in gaming engines seemed to be far way behind the graphics films. I'm obviously not an expert on these things, but what did they do wrong? There's always the "cgi/cartoon" look. The , the effects too. Is it the textures? Or did they let that happen on purpose? Our PCs probably can't handle all that, but every time they showcased a brand new engine, I always thought three things: 1. "Wait, didn't some movies had that kind of CGI years ago?"2. "What's so new about it?"3. "Is it just the ability to have that kind of graphics on a game?" I'm not complaining here, I'm just curious. I'm really happy if someone can explain. They obviously are working as hard as they can to improve the engines. Still, do we really need to wait for another Half-Life? Edited May 24, 2011 by Ancient Aeon Link to comment Share on other sites More sharing options...
jhardingame Posted May 24, 2011 Share Posted May 24, 2011 Well, the difference between CGI and what we have in video games is render time. For example, all the effects in Took 13 hours to render. A full CGI scene takes even longer. It's not that they're 'behind', it's just that you can't have CGI in a real time render, and run it on consoles/most consumer PC's. In all seriousness, having a "CGI" quality video game would lag worse than Crysis on a laptop from 2001 by a few hours, maybe even days. It's just not plausible to run something like actual movie CGI in real time and run it in a video game. Link to comment Share on other sites More sharing options...
AncientSpaceAeon Posted May 24, 2011 Author Share Posted May 24, 2011 (edited) Yeah, I knew that having that quality would need a super computer. The thing is, I felt that the hype for a brand new engine is far bigger than something like say, a brand new video card/development in hardwares. I could explain what I meant better if only I'm better at writing English. Edited May 24, 2011 by Ancient Aeon Link to comment Share on other sites More sharing options...
miscdude Posted May 24, 2011 Share Posted May 24, 2011 (edited) Wall of text warning. tl;dr: Ultra-realism in games is impossible. One of the hardest parts about rendering something to be photo-realistic in a still environment is light and how it affects things in a realistic fashion. Normal Maps have helped with squeezing more detail into fewer polys, and works great for things like reflection and refraction of light. But there are a variety of elements, most commonly on things like skin and reflective metals or water, which exponentially increases the complexity of the render and ultimately the render time. Making skin look realistic has a lot to do with Sub Surface Scattering, which is where the light phases through the skin, such as the redness you can see if you put a flashlight up to your finger. If you were to set up a high poly finger model in a 3d application and put a light next to it, and slapped on a regular texture, the light would simply be blocked. The render time for this would probably be relatively low depending on the rendering software as well as the poly count of the finger. Let's say for this instance, the poly count of the finger is 500,000 polys. Sculpted in Zbrush or Mudbox or something to achieve a very realistic appearance. Let's also say that it takes 10 seconds for the scene to fully render, finger and light-source. Now, you can reduce the rendering time by about 8 seconds if you utilize normal maps and drop the poly count to closer to maybe 12,000. So we're at one basic light, one normal mapped mid-poly finger. Now, you change the material used for the finger to a Sub Surface Scattering material (SSS.) After you set up the ratio of units to the in-program measurement of distance, you get something that resembles the red light showing through skin. The skin, however, looks like wax, and because of the complexity of the render going from 2D surfaces on a 3D object to making the 3D object have an assigned volume, the one light and one finger with SSS material has a rendering time of 30 seconds. That's not bad, not for a still image. But the lighting will soon start looking fake, and low quality when you tweak some settings a bit more. Then you have to adjust the sampling of the light to a higher value, 2x, 4x, 8x, or 16x what it was before. This will take the render time to lets say 2 minutes and 30 seconds for a somewhat high light sample value. Now you have your finger which looks very waxy, the light, and the red part of the skin from where the light is shining through it. Now let's say you want to add a high-resolution texture to the finger, add that little bit of life that isn't there with just a color and a shader. For realism sake, the texture size of the finger we will say is 1024x1024. You now have the diffuse texture, which will add about 10 seconds to the render, as well as the normal map from before. But things still aren't looking right, now you have to make a specular map and make sure that the skin is shiny in little speckles to emulate the texture a little bit better, and make it look less waxy. Now you make a specular map, which adds about 10 more seconds to the render time. Now we're up to 1 minutes 50 seconds. The finger itself looks alright, but something is off. That's because there is no environment, no other light, no other objects to bounce light off of to make the finger look like it fits it's surroundings. Now you add a studio box, just for the sake of rendering the finger, and you add two more lights with varying intensities and angles. Now, the render software has to render the box, the finger, and the three lights. This will take your render time, lets say, up to 5 minutes. Yes, it is that dramatic of a change(Especially with something like Vray. Oh my god.) So now you have one finger, one realistic looking finger out of 4 on a hand on an arm on a body. This still image takes 3 minutes to render. Now let's say you're animating it ever so slightly, just bending the finger a little bit, and maybe swiveling it in and out of the light. A very, very simple animation, with lets say 180 frames. At 60 FPS, the highest frame-rate the naked eye can see, the most fluid, and realistic thing you can actually look at, you're at a 3 second animation. Now here's the fun part, the reason that 3D animation is so utterly monstrous on time and why professionals spend thousands of dollars for high-end computers. That 180 frames, that 3 seconds of animation, at 3 minutes per frame boils down to 9 hours of rendering time. No joke. Sometimes you can optimize your settings so as to trim a few hours, but advanced lights (Realistic lights) and advanced materials with simulated volume (SSS) will take a long time. There is absolutely no way around that. Very, very few people who do 3D animations do them at 60 FPS and even fewer bother with realistic materials (See Pixar) v.s. photoshopping in effects afterwards. Now, I remind you, that is a 3 second animation of a finger and 3 lights in an animation program. It takes all of a computers GPU and CPU to pump out that quality of render, even having a multi-core processor doesn't help very much. Now, imagine if you had an entire scene. A full human body, simulated collision, cloth that moves realistically, hair that flows realistically, reflective properties in things like eyes, grass, dirt, trees, maybe some rocks. You're looking at an incredibly long rendering time for even a 3 minute animation. And again, this is all in a 3D rendering software - Something ENTIRELY devoted to making the graphics look good, not bothered by taking controls, calculating other inputs, just rendering information that is already there. Now, if you try to add this into something where the CPU is also handling things like real-time collision effects, skeleton-animations that can change at a second's notice. Then the GPU has to be powerful enough to handle basic lighting and texture application and the SSS as well as dynamic reflections, multiple objects casting shadows that need to be constantly re-rendered, a HUD, advanced lighting systems like HDR, Bloom, things like Anti-Aliasing all without much help from the CPU. It will be a very, very long time be impossible to achieve photorealism in video-games. The best that a game engine can really do is increase texture resolution which is already a hit on the GPU, increase the volume of things like grass and hair, refine the textures to a static perfection with their environment, and increase the amount of bones in the models that can move (also a hit on the GPU.) That is one of the reasons why certain games which boast about graphics (Such as the Final Fantasy series) rely heavily on cut-scenes. They're pre-rendered, and all it takes to show them after being packaged is a codec that can run the video file. "quality would need a super computer" is right, a super computer from about 15-20 years from now. Nothing as of right now is even close to powerful enough for the task, thousands of dollars are spent to reduce the 9 hours to 3 or 4 hours, not to the 1 or 2 seconds necessary. Edited May 24, 2011 by miscdude Link to comment Share on other sites More sharing options...
brokenergy Posted May 24, 2011 Share Posted May 24, 2011 (edited) This leads into the question; which is better photorealism or surrealism? This vid should explain what I mean. Edited May 24, 2011 by brokenergy Link to comment Share on other sites More sharing options...
Halororor Posted May 24, 2011 Share Posted May 24, 2011 I never quite understand this. Why would you want something that you use to escape reality, want to look like reality? Link to comment Share on other sites More sharing options...
JimboUK Posted May 24, 2011 Share Posted May 24, 2011 In time computers will be able to process at the speeds required to render like that in real time but we're a very long way away from that time. Link to comment Share on other sites More sharing options...
AncientSpaceAeon Posted May 24, 2011 Author Share Posted May 24, 2011 This leads into the question; which is better photorealism or surrealism? VIDEO This vid should explain what I mean.Yeah, I know about that. But that valley only applies to humans, it doesn't apply to other things. is an example, it could be called revolutionary, but there's something really wrong about their faces. What I really meant to say is how the graphics in gaming haven't got major improvements/breakthroughs lately. Have you seen the new ? Compare it to the MW 1 and 2, they look exactly the same, prove me wrong. , while the animations are outstanding, the rest of the graphics are just like a slightly better MW. It's like after they made Oblivion, Crysis, and other stuff, they just, gave up. Maybe it's done on purpose, though. X360 can only handle so much, and even PS3 are showing signs. And not everyone have a >$1500 PC. Do we really need to wait for Half-Life 3? Link to comment Share on other sites More sharing options...
HellsMaster Posted May 24, 2011 Share Posted May 24, 2011 ...Slightly better then MW...Check your sources again Ser. :excl: Also, something to keep on mind in movies is some of the movies are shot in real and then effects are added. Also it takes around 20m to render a Viper in 3dsMax using Vray so there is no chance that even the best of computers can do it.However i do remember a Nvidia quadro show case where it showed peopled real time rendering on a porshe.. and WOW :woot: Link to comment Share on other sites More sharing options...
JimboUK Posted May 24, 2011 Share Posted May 24, 2011 http://www.youtube.com/watch?v=Pyhf3JmODHE This is very impressive. Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now