========================================================================================= The main points from the roundtable: ========================================================================================= - Comparing the Xbox to the Xbox 360, and the Playstation 2 to the Playstation 3, CPU has increased from 16 to 64 times, but RAM has only increased 8 or 16 times. With all of the extra content being added into Next Generation titles (normal maps, specular maps, more animation, HD resolution cut-scenes, etc...) we must find more room to store it all. One possible technique is by using procedural content creation. - Procedural techniques allows for more content to be created and modified in a shorter amount of time. Does this mean that administrators will cut team sizes ? - Procedural content is only useful in certain situations, when it comes to game design: in the case of linear gameplay, this is less useful as it makes for a less controlled environment. However, in the case of open gameplay, it can help to fill in the cracks that take too much time to develop. - People tend to more easily forgive problems with procedural content in open games (ex: AI bumping into walls in GTA3) - In the case of SpeedTree, developers must be careful to strictly control or freeze the randomness of foliage generation, as it may block the player's critical path. - Quality Assurance... adding a lot of procedural content can be a nightmare for QA, since in some cases they have an almost unlimited array of possible outcomes, as opposed to a more linear and static game. New tools will have to be developed to test this (i.e. AIs to test AIs). - Diablo and Diablo 2 had procedural level design... can level designers justify their jobs if procedural level design becomes so mature that it is as good as we are ? - Adding random noise to EVERYTHING, not just textures or animation... dialog, the environment, AI behavior, etc... - A lot of procedurally generated missions in games (i.e. Anarchy Online's Mission Generator) seem inhuman and hollow. Can they be sophisticated enough to seem as good as something created by a designer? - Procedural content software is not simple to use. (i.e. Houdini, etc...) Artists have to become more and more technical. - Many games have huge expansive procedural worlds... but is the gamer going to want to, or have the time to explore everything ? ========================================================================================= Multimedia files were used to show procedural content in a variety of different forms: ========================================================================================= - Will Wright's Spore video from GDC 2004 is well known as having kicked off a resurgence of interest in procedural content generation. http://www.spore.com/ - Genetica is a software that allows textures to be created procedurally (although the final product is a bitmap, it is an example of what a run-time procedural texturing tool could look like) http://www.spiralgraphics.biz - KKreiger is a first person shooter on par with Doom 3 (stencil shadows, bump and specular maps, multiple weapons) which is contained in 96K. When the game is launched, it unpacks to 275 Megs. It is able to do this because all of the content is procedural. http://www.theprodukkt.com/ - SpeedTree was used in a few recent games, most notably Bethesda's Oblivion. It allows flora and fauna to be generated from a very small file containing simply variables that decide the type of tree, how it will grow, etc... Not only does this make the trees easy to modify, but they take up much less space on disc. www.speedtree.com/ - Endorphin is an animation software that combines three different techniques: Keyframe animation, ragdoll and behaviors. The first two are well known, but the last is a sort of AI that allows procedural animation to be created, thereby making the job of the animator easier but also reducing memory and disc space. http://www.naturalmotion.com/pages/products.htm ================= Procedural Asset Creation ================= Static and Dynamic ================= Static: Has been possible for a long time... - Prepared before time and stored on media (Not talking about this) - Prepared before time and generated before game start Dynamic: More possible with next-gen hardware - Done completely real-time Next-Gen Systems... more CPU, but less RAM ! CPU RAM Xbox: 733Mhz 64MB X360: 3 x 3.2GHz (9.6Ghz) 512MB Diff: 13X faster 8X faster PS2: EE + GS (346Mhz) 32MB PS3: 7 x 3.2Ghz (22.4Ghz) 256MB Diff: 64X faster 8X faster Limited media size... BluRay: Single 25GB Double 50GB HD-DVD: Single 15GB Double 30GB HD Cut-scenes: 3X larger than NTSC Like the introduction of technology, will this remove jobs from the industry ? Will artists be replaced by technical artists or art director / shader programmer combos ? Geometry ================== http://gsculpt.sourceforge.net/ - SpeedTree - Houdini: Everything is a node, you can plug XYZ values into RGB values, auto converts for you, etc... - Dynamic generated dungeons in Diablo Textures ================== - Shaders - Genetica: http://www.spiralgraphics.biz/gen2tour/tour1.htm#tour1 Animation ================== - Building dynamic animations - Naturalmotion Endorphin Audio ================== - Adding noise / randomness (already done in certain games) - Does emulated audio on PS3 and X360 change things ? Games ================== - Spore - Microsoft Flight Simulator - .KKreiger http://www.theprodukkt.com/