Slashdot Mirror
Carmack GDC Keynote Rambles Fascinatingly On Re-Use
Thanks to GameSpy for its coverage of id co-founder John Carmack's keynote speech at Game Developer's Conference. Ideas discussed include the now-dismissed concept to do a 'Quake 2 remix' - "to rebuild the game using all-new assets and technology", as Carmack pointed out: "...even the idea of just reskinning an old game brings with it the problems that as we have newer graphics technology, media creation demands get worse and worse." Of follow-ups to DOOM 3, Carmack also mentioned that "they were hoping to re-use some of the assets created for DOOM 3 to help speed up development of whatever they do next, but even that would have a tradeoff."
In general, I'm all for remaking classic games... but Quake2 isn't one of 'em. People are already grabbing the old code and makeing not-for-profite rebuilds of the venerable older iD engines, such as the tenebrae project on sourceforge for Quake1.
I'd love to see some remakes or re-releases of various RPGs and console games, but to stay afloat the PC Market basically needs to be constantly defining the upper limit of graphics, AI, sound, interface, gameplay, and every other "aspect" you could assign to videogames.
That's where I've always seen iD as shining anyway. They're responsible for bringing shareware and 3D gaming to the mainstream, back in the day. They made vastly popular "true 3D" graphics (yes, I know there were other true 3d games, hell, I loved Descent and that one was perhaps MORE 3D) introduced colored lighting to the world, did some neat things in Quake 3 with texture effects... I buy their games so that they can keep making the engines that modders and other game developers then turn into fun products. And that's the niche I'd personally like them to continue to fill. Make a good engine, at the higher end of the technology curve, that others can play with.
Switching gears, Carmack touched on the subject of audio in games, saying that if you wanted to spend all the computing power currently available, you could do just about anything with audio in games. The question at the moment is whether it's worth devoting that much effort and horsepower to audio. "It really doesn't pay off in the current generation to put that much effort to it," said Carmack, "but give it a couple more turns of processor generations, and audio will be just be 'done.'"
IANAGD, but this seems short-sighted to me. The state of audio in games nowadays is incredibly primitive. There are a limited number of sound effects and every time an event occurs you always hear the exact same thing. This is like saying that since the sprites in Mortal Kombat are photorealistic, graphics will never improve much more than that.
A truly advanced audio system will dynamically generate sounds that are uniquely appropriate for the corresponding event. So when an object falls into a body of liquid, for instance, it takes into account the object shape and size and speed and density and the viscosity of the liquid etc. The Half-life 2 engine seems to do more or less do this, but it's not out yet. And I think it just calls upon a large database of sound effects rather than generating something truly unique.
Another thing is that right now every single line of speech in a game has to be recorded and stored, which a herculean effort for a large RPG, and generally limits the possibilities of conversation. Hearing the same stupid bark every time you encounter a character in Deus Ex 2, for instance, is badly grating. Ideally, designers will be able to just write dialogue (or better yet design an AI which dynamically generates dialogue), add a few parameters for inflection and rythm, and have a text-to-speech engine take care of the rest. Considering how non-real-time state of the art text to speech is laughably bad, I don't see how throwing more computational power at the problem is going to solve it.
And then there's the problem of creating music which reflects the onscreen action. Just looping the same music over and over works OK if the music is well-written, but ideally it should be responsive to what's going on in the game world. This isn't just a technical question - how do you even compose music or bits of music that flow together seamlessly and in any order? An then how does your engine call upon these bits of music? Or do you create an algorithm to randomly generate music? In general, random content generators, whether level design or art or music doesn't come anywhere near the level of quality of hand-crafted material, at least for any algorthm that I've seen.
All that said, I do believe that Carmack's main point that the cost/benefit of developing an advanced audio engine is not nearly as important as graphical innovation. People notice graphics much more than audio and you can get away with a relatively crude audio system if the graphics are pretty. Also a good audio engineer (like Eric Brosius who did the audio for the Thief games and SS2) who creates fantastic music and sound effects is far more valuable than highly advanced technical engine. I just think that there are a lot of serious technical and artistic innovations that can be made, even though they may be subtle and initially underappreciated by the average user.
if you wanted to spend --all-- the computing power currently available
note: he meant that if you could put every single clock cycle into audio of a current high-end CPU (no game, no graphics, just audio) you could 'solve' most things. While there is really no limit to how much computing power you can sink in fluid dynamics I have the feeling that creating approximate sounds dynamically would not be as computationally expensive. Games nowadays probably put only 1% or so of the CPU's computing power into audio I believe...
John also says talks about give it a couple more turns of processor --generations-- , this doesn't mean the 100MHz stepping changes that Intel/AMD are doing, this would mean the same difference in computing power between a p4 3.2GHz and a pentium 1 133MHz (3 generations roughly), I think it's fair to assume that if you had a p7 CPU clocked at 60GHz you could afford to devote enough computing power to sounds to be able to create fairly realistic synthesized sound effects for pretty much everything, not to mention fairly believable voices while having still tons of horsepower left for game logic, graphics and AI.
-- the cake is a lie
Carmack also mentioned that "they were hoping to re-use some of the assets created for DOOM 3 to help speed up development of whatever they do next, but even that would have a tradeoff."
How about lending some "assets" to 3Drealms to help speed up the development of Duke Nukem Forever. God knows, they're due for another engine switchover.
Uhh.. I daresay half-life + mods is vastly more popular than quake3/arena..