I'm not going to weigh in too heavily on one side of another here, but I think the typical argument used to combat this is perceived quality versus actual, technical quality.
I already pointed that out so it isn't my point at all. It is a given that the PS3, nor any other current console, is even close to creating the effects to the technical percision seen in the CGI.
The question I have posed is what techniques can be used to get the same visual quality within the game experience. It doesn't need to be the identical technology or technical quality (i.e. full, accurate simulation) but they do need to be effects that are not perceivably different in the game experience. For example, if you have static geometry in a CGI using GI, you could possibly recreate the look in realtime using a precomputed radiosity solution. It wouldn't technically be the same, but in regards to the output on the screen it would be near impossible to distinguish if the constants (static geometry) remained unchanged.
That is an example to get CGI level quality in a game without doing all the hard work an offline renderer does. Likewise there are times when normal maps, and texture maps, can be selectively used in ways to mimick real geometry with the viewer being none the wiser. e.g. In the ToyShop tech demo they went to great lengths to create very nice parallax mapping, and the designed their assets in a way to hide the situations that destroy the illusion (e.g. putting cornered hedges on the buildings so the user could never see the wall flush, which would destroy the illusion of real geometry).
So I absolutely agree that there are ways to perceivably get CGI quality in a game without actually recreating the CGI technology or quality.
The question is how long until we get KZ CGI quality in a game, and as the KZ CGI stands, there are far too many places where hacks are not going to suffice to create a realistic illusions that won't demonstrate a lot of disparity with the CGI. IMO it is far, far to complex (and long!) of a video, demonstrating far too much polish, detail, and quality to be cut corners with in a way that doesn't detract from the quality goal.
If there are, I am excitedly awaiting suggested techniques to deal with the number of issues I and others pointed out from the video that are non-trivial hurdles.
A great artist or set of artists might achieve a lot with much less technical resources..I guess the question is if that level of technical complexity is required to have something appear to the majority to be reasonably "level" with the KZ movie or something of its ilk.
Majority? That is a significant shift from what I or the OP were discussing.
The question was when would we see KZ quality in graphics -- the OP (Rangers) was wondering if it would be the PS4 generation. To that inefficient bluntly stated:
inefficient said:
It will come this gen.
There isn't anything particularlly impossible about it minus the bullshot AA.
He says it isn't impossible -- as a statement of fact -- and yet refuses to give any substantial evidence to support this opinion. This is the source of the disagreement, not the issue of "perceptions of the majority".
I don't know myself, but I guess that's where the uncertainty and speculation begins, as opposed to the known closed-book case of whether PS3 could apply the exact rendering techniques used in that movie in realtime (which it can't).
I would agree that the trained eye will never be as pleased as the masses might end up being, though, for sure. That's part of the problem with this argument, though, because it will come down to perception which is entirely subjective.
KZ CGI quality isn't subjective at all. Either it is, or it isn't.
Back to the majority point, if all you had to do was fool the majority then mission accomplished already. I have already seen people confuse NBA 2K7 for a live basketball game. If the issue was the perceptions of the majority, this is how I would go about doing KZ (I am sure there are other approaches, this is just my suggestion how to get close enough that most casual observers not scrutinizing the image won't really notice the differences):
Joshua Luna said:
- Filtering and texture quality should be the main priority on the rendering end. Textures need to be very clean and remove as much shimmering and nasty dithering as possible. 8xAF, if not 16xAF, should be used as often as possible as well as any close up objects should have extremely detailed textures. Keeping the "grey" them can help in this regards with a focus on materials that interact differently, creating a high perception of object detail.
- MSAA should be the next major render focus. 4xMSAA minimum, as well as alpha channel AA. With AA and quality texture filtering the image will look very clean. It won't be anywhere CGI quality and will have artifacts the trained eye can see, but it should be a step up from many "next gen" games and removes huge eyesores from the "majority" perception.
- Normal maps on the characters and vehicles. Normal people can hardly tell the difference between a normal map and real geometry.
- Volumetric clouds. Rip off the Warhawk code. The opening could be done with a mixture of realtime, raycast clouds and a static backdrop of fake textured clouds. Most people would never know the difference, or even care.
- Hair. Something has to be done with the hair. Stealing the approach Gears of War used would be a good start, although this is one of the areas I think would just stick out, especially when comparing the shockwave burst. But GOW style hair would fool a lot of people, at least initially.
- Particle system. You will want to save a lot of fillrate for the particles. They also need to be very dense, well lit and shelf shadowing as well as casting shadows. We have seen some techniques that create similar (not identical) particle clouds in realtime, and Crackdown has some similar explosions. We would need to aim a couple steps above Crackdown in regards to particle density and look, as well as make them particles stick around longer. Particles would be a major component of your design as lacking particles would be one of the first major eyesores. Anything that moves and interacts with the environment needs special attention and focus. The slow dissapation of smoke would need to be a focus.
- Vehicle destruction. Goes hand and hand with the particle system. Motorstorm and Burnout Revenge have shown you can do some nice vehicle deformation and destruction on a large scale, for a handful of vehicles should be able to be done on an impressive level. A lot of detail needs to be worked out here, though, like the shocks on the car on the bridge and getting realistic explosions from the vehicles. Probably one of the more achievable areas of the CGI in regards to getting enough quality that people don't care, at least compared to the KZ CGI.
- Lighting and Shadowing. The environment has a bit of destruction going on, so the best you could go far is clean shadow edges (soft shadows) and nice dynamic shadowing. A robust lighting and shadowing system can fool a lot of casuals, but we are very, very far from fooling those even passingly interested in graphics. Even Crysis, which is robust in these areas, has visible artifacting at times. You could probably even cut self shadowing (see the inbound flight for some nice examples) and most casuals would not notice. As long as shadows are clean and possibly have some sort of ambient occlusion (or somilar) as well as your normal maps the system should be robust enough, with all the other features, to give the general appearance to the majority.
- AI and Animation. They are very different topics, but this is one of the areas where the CGI absolutely sets amazing hurdles. The commanders face scrunches with the explosion goes off, you get handed a gun from a squadmate when you are not even looking at him initially, people jump from vehicles when they see they are doomed, the AI players not only communicate in regards to adapting to the battlefield (insanse scripting) but also use subtle and overt communication on the battlefield, your player takes a slow 'realistic plodding" walk up each of the stairs with the bouncing camera staying focused but showing the inertia, AI have a number of "exit the vehicle" animations, seemlessly dynamic death animations where dieing players react to the wound and to the environment (ala some of the Endorphine style stuff where you have a mixture of mocap and physics driven animation), and on and on. As mentioned above, anything that moves and interacts with the environment is going to be the first and last place that make or break your CGI quality attempt.
- View Distance and Detail. LOD is going to be your friend. Obviously a dozen 1M poly characters isn't going to happen, but even normal maps will have to be designed well for gracefull LOD. The CGI shows insane drawdistance with a ton of quality and detail in distant objects. Obviously with a game you are going to cheat some with Motion Blur and DOF -- and the majority of consumers will even give you a couple extra kudos for looking better for doing so! -- but LOD is doing to be an intigral part of the deception and illusion. Packing as much detail and quality into objects when they dominate the forground is going to be vital. The second you look at a wall from 2 feet and see low detailed, poorly lit with jagged shadows crawling all over the place with horrible bilinear filtering it will become unmistakable you are dealing with a game falling miles short of the CGI.
- Music and Sound. One of the key elements of the CGI is the quality, variety, and 'depth' of the sound. It sounded like a battlefield recorded on a battlefield. It didn't feel like dozens of sounds just tossed ontop of eachother, but the sound enhanced the visuals. Great visuals and sound can be found in the same game (see Gears of War), so this will be an area where a lot of elbow grease needs to be applied. It is like food: Good looking food enhances the flavor of the product. Making a game sound good enhances the impression of the game and reinforces the visual experience.
So, imo, many casual viewers would take the above product and have a hard time seeing, or caring, about the differences.
But that was never the original question and doesn't really apply to my responses to inefficient. Because, afterall, there is nothing necessarily impossible to do on the PS3 in the KZ CGIs other than the AA
I would echo inefficient's point about expectations at the beginning of one generation versus results at the end. I doubt anyone expected we'd see some of the stuff that we did in the previous generation, even if lots of trickery (or..intelligence, as some would prefer) was applied to bring that about.
I don't think this logic applies, as mentioned before to inefficient. Look at it as a logic equation.
1. The PS2 far exceeded our expectations.
2. The PS3, we can assume, will exceed our expectations.
3. The PS3 will produce KZ CGI quality visuals.
Just because #1 is true, and #2 can be assumed to be true by the nature of the industry, doesn't give us any reason at all to believe #3 is true or will be true.
Logically it doesn't follow. The fact the PS3 will exceed our expectations doesn't equate to the PS3 producing CGI level visuals.
If that were the case, could we not argue the PS2 will obtain CGI level visuals? At least to the majority?
When the question is altered to be subjective it ultimately breaks down. I can remember my parents claiming Mario 64 looked like a CGI movie (Toy Story). Personally, even if the PS4 doesn't reach KZ CGI quality it won't matter IMO because by the time the PS4 reaches its peak in 2016ish the differences just won't matter in regards to gam mechanics/gameplay/immersion and creative content. I think fewer and fewer people will be able to see the differences, and even fewer will care.
So I am not poo-pooing on that general idea, only that it doesn't apply to the current discussion.
edit - I've just noticed I'm basically skirting around the same arguments you both have already made, but I guess I'm trying to say you're both probably right, just occupying different perspectives
Not really. inefficient believes nothing but the AA is impossible this gen and that such insane elements, like the animation is "totally doable" this generation.
I am saying that, at best, there will need to be major compromises that have many areas where the user "breaks" the experience. And that is taking into consideration and assuming artists are able to come up with many implimentations that in best case scenarios rival the CGI.
And of course I am willing to give examples and listen to technological reasons (e.g. I listed some techniques that could give some nice results that would suffice), and, well, the other side is more argueing out of belief.
Of course as of Fall 2005 over 30% of users who voted here at B3D believe KZ PS3 will look
exactly like the E3 2005 CGI. There is still a large undercurrent of folks who believe it was real and think Laa-Yosh and others are liars, so there are plenty of people who want to buy into the belief it WILL happen. ::shurggs:: I don't really personally care, although it grinds on me to see places like IGN call a game like Motorstorm looks like the CGI. Because it doesn't, even to a casual eye. Unless, of course, you want it to
But then again, most people are looking at the art and feel of the game and not the visuals. And I am not talking about the visual quality, but the visuals in general. If the artists capture the general artistic style, but fall short in regards to quality, people have a hard time seeing the difference.
The same reason my parents though Mario 64 looked like Toy Story.
