Sponsored By


Opinions about how procedural generated content (from NPCs to buildings) can minimize work and development costs. (PART I of III)

Boto Gatas, Blogger

September 7, 2010

4 Min Read


If I was a developer, especially one that creates persistent open world sandbox games, I would invest in a tool that could randomly generate game content, starting with NPC characters (much like the Sims 3 character creator, but automated). This tool should be highly configurable (for instance, if you want a stage in the game to be filled with fat bald guys, you would limit the hair and weight variables to adequately match these characteristics). Also, once the character was randomly created, you could fine tune it to fit your intentions. Other instances could be incorporated, say, to generate zombies, with randomly generated wounds, exposed organs, bones and the like. Such a tool, in my opinion, would greatly reduce development costs and add variety to a universe within a game.

Along with the aesthetics, the tool could incorporate movement and behavior variables, for example aggressiveness, speed, weapon carried, stamina, items dropped and so on, as well as the possibility to create models of correlation between such variables. These variables could be also be linked with the aesthetic/physical attributes (e.g. fat guys run slower). Thinking far ahead, the tool could have a totally configurable AI routine, with a huge library of prescribed actions and their variables, so the developer could select what actions a certain group of characters should perform, the range each variable would oscillate and let the tool generate the characters along those prescribed lines. Of course, new actions could be developed and added to the library, ever enriching the AI capabilities.

I believe random generator tools to be built with future applications in mind. So they should have the capacity to create Hollywood caliber graphics as default and use some kind of algorithm to resample those models so to fit each specific hardware processing power. This way the developer could create one “utopian” highly detailed model and export it as a lower polygon version to be used, for example, on a PS3 game or an even lower version to a Wii game. This way the tool would be always up to date to technology trends.


The same way, but in a more open-ended style, an animal/monster generator tool could be created, like Spore’s, but with way more freedom. It could be integrated/mixed with the people tool generator in order to create human-beasts characters. The same way as Spore’s engine, the tool would generate the movement behavior of each being, based on their skeletons, plus weight. It could and should have a library of AI routines as well. It could be configurable to only generate beings that are physically plausible. Or not. The interesting thing of randomly generated creatures is that, most certainly, beings would emerge that no one would be able to imagine, fantastic absurdities proportional to level of freedom given to the variables.

Once again, even if the developer desires to create the creature he/she envisioned, instead of modeling from scratch, he/she could select from the libraries of body parts and adjust its variables so to fit his/her vision, which I believe is a faster and easier process.

Trees and plants

Trees (and plants in general) are modeling hurdles that should long be overcome. I think it’s much easier to create a random generator to them since there are fewer variables, principally regarding movement. It would be cool if the plant generator could be mixed/integrated with the beast generator, in order to create hybrid plant-human-animal beings.


It’s even easier then plants, but I think that by being a so fundamental part of the a game playability, creators will ever want to have total control about it. This does not impede of incorporate a random terrain generator, of course.


If I was a developer, I would create a way to import CAD plants to my engine and buy as many of them as I could. If I had a team of really smart programming guys, I would ask them to develop a system to analyze each of these plants I bought and learn from them and, applying physics, generate random buildings based on what was learnt. I believe a human guiding hand would be necessary (at the early learning stages especially), in the form of input about what is right and what is wrong with each generated plant. I don’t know. Maybe this is too outlandish. Maybe a simplified building modeler, like, again, the one in The Sims, would be the easiest solution.

Read more about:


About the Author(s)

Daily news, dev blogs, and stories from Game Developer straight to your inbox

You May Also Like