The Necessary Fantasy Logic

In my second post about videogame architecture over on BLDGBLOG, I connect some ideas about the design of videogame cities to the possibilities of procedural generation. It’s a little bit speculative, especially when I start to link Half-Life 2’s Viktor Antonov (whose concept art for The Crossing is above) to Chris Delay‘s methods in Subversion:

“What parameters would we need in order both to understand and create a science fiction Edinburgh, or Sao Paulo, or Vancouver? Identify the necessary fantasy logic within a procedural city-building system and you could recreate cities with their alternate identity in an instant. An accelerated future Moscow, or a retropunk Venice, instantly sprawling out of the monitor.”

Stuff like that.


  1. windlab says:

    These are the important things in gaming.
    Not painstakingly created environments, but environments which painstakingly create themselves.

    Unfortunately I cannot see them being realised soon.

  2. Quine says:

    That Structure app looks amazing. Can’t find any shots or footage of it so far though.

  3. Jim Rossignol says:

    Structure was removed from the web a few weeks ago. Presumably someone bought the tech and vanished it.

  4. Ginger Yellow says:

    So, Jim, and anyone else who has thoughts on the matter, what do you reckon is the big picture consequence of all this procedural content generation that’s being worked on? Assume that in five years or so developers have a fairly well established set of tools for doing procedural animation, procedural textures, procedural landscapes etc. What does this mean for games? Will it be used just to cut costs? Will the money saved on content creation be spent on, say, AI programmers? Will it actually open up new types of gameplay or will it just be a different way of creating the same games?

  5. dartt says:

    I found this video of Structure on youtube:

  6. Jim Rossignol says:

    I guess you could generate entire cities, then go in and tweak it around your game, rather than building the city for the purpose of the game. It might mean that large environments are more feasible, with level designers becoming modders and polishers, and only building very specific pieces of scenery.

  7. Mike Arthur says:

    @Jim: IIRC isn’t this what was done with Far Cry 2 (but without the cities).

    I definitely think this stuff is the way forward. Nice link!

  8. ourdreamsoffreedom says:

    In the future you will boot up your XBOX3000, select which game genre we want to play and five minutes later you will be playing a generic randomly generated game.

    There won’t be any more individual games, only packages of environments, characters and equipment.

    Only I won’t be around then. I will have long moved to the woods to live with the squirrels, far away from the mechanical precision and pharmaceutical drugs that neutered our creativity.

  9. Carey says:

    I’m with him. Squirrels are tasty.

  10. Jim Rossignol says:

    Technology doesn’t neuter creativity. Luddite nonsense.

  11. Kieron Gillen says:

    Totally. Technology has only ever facilitated creativity.


  12. noony says:

    I think the real power with this concept is not that developers could use it as a labour-saving device, but more that the players themselves can have control over the creation process. The developers’ role would be in producing the theme and context of the game, but players themselves could fine tune parameters which decide the ultimate appearance of the environment or the complexity of AI behaviour.

    Imagine something like Spelunky or Dwarf Fortress, but that generated levels of the quality in Far Cry 2 or GTAIV. Then imagine that the game kept track of how long you spent in each environment and how successful you were, and then generated levels that it thinks you would enjoy ever-so-slightly more. This sort of thing has been looked at in academia and I don’t think we’re particularly far from seeing it utilised in a commercial setting.

  13. dhex says:

    Technology has only ever facilitated creativity.

    welllllllllll, not entirely. being in the uk surely you are familiar with the horrid mess that is euro techno? :)

    but in general, yeah. this kind of generation saves time for individuals or teams to focus on design. that’s good news indeed.

  14. Kieron Gillen says:

    Technology also facilitates a lack of creativity.


  15. Nuyan says:

    You could however make a convincing argument for saying that technology neuters imagination and fantasy.

    But interesting articles you got there on BLDGBLOG. It’s the sort of stuff about gaming that I find fascinating.

  16. Jim Rossignol says:

    “You could however make a convincing argument for saying that technology neuters imagination and fantasy.”

    I don’t think it would be as convincing as an argument saying the opposite.

  17. Theory says:

    Is The Crossing really cancelled? Last we heard it had been shelved until Arkane were in better financial shape.

    Also, Dartt’s video is incredible. But without a human spark behind the design of a place, I wonder how long it would take for exploration to become depressing.

  18. Jim Rossignol says:

    That’s why a hybrid approach would be the way forward: use the tech to generate the city, then have level designers tweak the produce into excellence.

  19. Mark says:

    Being a videogame environment artist for a living I can’t help but look upon this with a bit of trepidation.

    Tools like this will open up large city environments for use by smaller dev teams in and smaller games, which is great. The high time / money cost of art assets does limit smaller projects in size and this will be liberating for them. A gifted programmer can code half a game (certain types of games) by themselves, a lone artist can’t produce more than a certain amount of content (especially next gen contant) no matter how good they are. Personally in the future I can see the possiblity of seperate companies that purely provide environments for use in games simulations etc, instead of each software developer making their own. Instead of making your own new york you can simply buy or license one and go from there.

    At the point when graphics are the same as reality/photography or get so close that changes in realism of an image are negligable, I think we will see a new period where content finally becomes cheaper to create/buy. Once someone has made an asset, building, city etc in full detail, it won’t need to be done again.

    Regardless, textures and art assets will still need to be made, and any city / game with any artistic style rather than a bland computer generated landscape will need plenty of hand tweaking or custom assets. Although shaders may eventually reach a point where textures become moot, and all detail can be represented by a material/shaders properties when combined with a super his res mesh or disp/normal/height map equivalent.

    Very cool but rather scary stuff. I think most artists are safe for a while though, people that want cities with a bit of character or soul will still be making or heavily modifying their own for a few years yet. Thats the artists that haven’t been outsourced to china/india i’m talking about.

  20. Baggy says:

    So if at a point the technology gets accurate and, well, good enough, the development team would simply supply the story, missions, AI, textures, and models/characters, and then cities to hold these stories could be generated on the fly, textured accurately and believably, plot points and missions could be algorithmically placed in such a way that made sense. Every player could have a unique environment to experience the game in. Every playthrough could generate a new backdrop to the game. This seems fascinating to me.

    Although it would ruin GameFAQs forever.

  21. SanguineAngel says:

    I recon you could probably argue that whilst technology allows for greater creativity, it has been the general trend (though there are those that go against the grain here) that alot of developers just use higher tech initially as a crutch and this is usually to the detriment of creative content and design.

  22. Noc says:

    I am wondering, though, how quickly procedural city generation will become . . . I guess “transparent” is the word? Where we’ve become familiar enough with the particular generation engines used that instead of seeing these fantastic, complex environments, we just start to see the quirks of the algorithm.

    Like with physics or rendering engines or such, when you reach the point where Source games end up “looking kind of Source-y.” When, even though Game Engine X renders things in an immensely complex and intricate way, the product quickly starts looking very familiar, and it remains a challenge for a design team to produce something that doesn’t bear the telltale sort of “watermark” of the engine.

    And I’m wondering how quickly we’ll start to see the same sort of intersections and the same sort of permutations on building shapes and park borders and bench placement, and instead of seeing a vast, ever-changing urban landscape we’ll sigh and think “Oh, they’re using THIS generator.”

    (Or even, within the same game, for one city block to start looking a bit like the next. Though it’s not as if that doesn’t happen anyways.)

  23. JonFitt says:

    Procedural generation for the laborious large bulk of things, and an artists eye for tweaks and the places and items of interest. That’s the best.
    You don’t want to pay someone to place all the trees and rocks and smooth every hillock in something like Far Cry 2, you’d want it to be generated, and then walk through correcting weirdness and adjusting things for aesthetic and gameplay reasons.

  24. Noc says:

    @Mark: Yeah, I don’t think the part of the artist’ll go away with or without, this.

    I mean, even if procedurally producing photorealistic renders becomes trivial, you’re still going to need an art team to produce a product that looks unique, and fits the game. (And is a well designed level, etc.)

    The metaphor that comes to mind is all the “functional” art fields: Architecture and Clothing Design and Industrial Design and so on. You don’t need to work to make a wooden house look like it’s actually made of wood, but you still need to put a fair bit of time into designing a house that’s aesthetically pleasing and worth living in.

  25. Alex says:

    Vancouver is already a sci-fi version of Vancouver! It looked awesome as Caprica City.

  26. Bobsy says:

    I have a special beserk button reserved for people who dismiss procedurally-generated game worlds as being dull, formulaic and unartistic. This is unquestionably the most exciting field of game development for me, and has utterly limitless potential. The ludicrously over-ambitions Infinity is a prime example of what the future of procural generation can offer.

    One thing that bugs me about game cities is that they’re never cities. They’re TINY. The entirety of Azeroth, famously, is about the size of Newcastle-upon-Tyne, a medium-sized city in the north of England. This means that the various in-game towns, villages and other settlements are shrunk to miniscule sizes. Source material for the pen and paper WoW RPG describes Ratchet as a bustling city-port, but the sad reality is that it’s got about five seperate buildings and a two-figure living population.

    Procedural generation makes for BIG cities. Cities that you can actually believe could exist. Cities you can lose yourself and spend hours, days exploring. I really want that. And it makes me go all BOBSY SMASH when I hear people whining about wanting smaller, hand-designed locales because they assume that’s the only way it’ll be interesting.

  27. JonFitt says:

    @Bobsy. The problem with real cities is that they’re full of uninteresting guff.
    People think they’d like a really large game world, but if you just want to get to the Opera House to save the girl and escape via the docks, driving for 20 minutes round the bypass and then spending 5 minutes stuck behind a bus outside WHSmiths, would drive you mental. Like in real life.

    Also, how far do you go with it, is the WHSmiths a real shop to go in or a texture on a wall? The fully procedurally generated ‘complete city’ route says that it’s a real shop you can go in, with a real staff room, and a real working staff toilet. But none of it is in any way relevant to getting to the blinking Opera house.

    I would like detailed explorable universes, but I think in reality I’d get bored looking for the interesting bits. Films and Theatre skim over the mundane parts to get to the good bits, and I think games do that by condensing spaces.

  28. Nighthood says:

    A good procedural generation engine has almost limitless potential, but carries with it a small problem. Games like GTA4 are good because you gradually get to know the ins and outs of the place you are driving through, but if it’s randomly generated, you’ll never get that feeling of knowing where everything is, unless a file is kept that notes everything you have seen so it can put it back there, which really isn’t very practical. The technology might be there, but the gameplay mechanics aren’t.

  29. Marcin says:

    Wow, that Structure video was very impressive. Throw in good car mechanics, some traffic and pedestrians and I would be easily amused just by that for months, just making and remaking cities and exploring them.

  30. Lack_26 says:

    I would love a procedurally generated Flight sim.

    But anyway, I would love a piece of software that allowed you to generate your own city and then change the different styles of the buildings, eg.

    Style: Old-town: High Gothic. New areas: Neo-Classical. Amount of Arches: High.

    Or that sort of thing. (And then preferable let you smash it with Nukes).

  31. JonFitt says:

    Procedural generation doesn’t have to mean random. Elite used procedural generation to produce it’s star map, but it was always the same.

    I could have a procedure to place trees every (2n+1) metres. That’s easier than making a list which says 1,3,5,7,9… metres, but every time I run 2n+1 it’ll give me the same values.
    I can design complex formulae which produce really chaotic looking patterns, but it’ll be the same every time you calculate it. Tweak a value and the output changes to something completely different. All of this without random numbers.

    …and that’s without even going into the fact that computers mostly don’t do really random numbers anyway…

  32. El Stevo says:

    “the cryptic (and as-yet-unexplained) Subversion, of which little is known, other than it relies on large, procedurally generated cities for the backbone of its game world”

    Subversion is a hacking game, or so Mark Morris said at a talk I went to. Obviously it has hacking in it (Chris Delay has talked about it on his dev diary), but the understanding I got was that that’s that main focus.

  33. Shadders says:

    Excellent article! And excellent links too!

    The once removed link to Eskil Steenberg was really interesting. (link to His development tools look really stunning.

    @Mark, I think your artistic abilities are safe for the near future at least. Procedural generation provides a means of creating content that tends to be more generalised. A specific asset still needs a human touch. Would the film Alien be nearly as good with a procedurally generated alien? It took the genius of Giger create the alien, the ship and the terrain of the planet to help that film become the classic that it is.

    It’s truly amazing what can be created with the ability to add, subtract, multiply and divide though.

    I really wish there were a few primers on how to get started with this technique though.

  34. Daniel says:

    Procedural deployment of cities has been around for a while with games. I can remember the countless hours I sank into the sim city watching the game interpret my intervention into a workable landscape. It would take my designations of residential, commercial and industrial and build be a scalable world.

  35. Jacques says:

    Great article, I do love BLDGBLOG.

    Procedural design seems to me like a good thing, it gives artists the time to spend time on the small, important details.

  36. underproseductor says:

    Procedural generation is a great way to reduce both making time and expenses, but there’s still a need for artists and designers to do some final touches, put in unique content and change some things. Still, it can save you the dirty and boring work.

  37. LionsPhil says:

    I really wish they’d rename that. Don’t they know that there is already a major piece of software called Subversion?

    What’s more worrying is that it might mean they’ve never heard of version control, yet develop software in a team. Shudder, etc.

  38. Kadayi says:

    As someone in the architectural/urban design profession the downfall I see to procedural design is that, Cities don’t appear overnight, they evolve from earlier settlements and its generally the case that short of massive degrees of deliberate restructuring (as in the case with Paris under Baron Haussmann), they will generally retain a degree of character from these earlier times that the more modern aspects of the City will build around. It is rare if ever that the past is totally swept away, esp when it comes to historically important buildings where the wheels of bureaucracy never cease to turn. It is only in Post Industrial Cities that uniformity of development is generally found.

  39. Jacques says:

    I think Kadayi may be referring to Milton Keynes.

  40. Jambe says:

    I’m a conceptual artist myself, and environments are my favorite things to work on. What I find most interesting about the notion of procedural building generation is the question, “How will it interact with natural features?”

    Say you’ve got the landscape already β€” a river twisting and turning through low, rolling hills β€” and you want to generate a believable city centered around one of these bends (as in the second-from-top screenshot in Jim’s article).

    How will the generator decide which buildings to place right along the waterline? Will it construct bridges? Of what type, and in what locations? Piers and docks, perhaps? What about long, winding avenues of fancy-schmancy private residences? There seems to be much focus on cities in this article β€” who’s to say you can’t generate a suburban or, hell, even rural sort of community?

    This is all very, very interesting to me.

  41. jalf says:

    @Kadayi: The article sounded like at least some of them tried to take that into account and build the city gradually, over time.

  42. Dave L. says:

    @Theory: That’s the last thing I’d heard about The Crossing, too. Arkane seems to have pulled all mention of it from their site (other than an art shot in the Jobs page), but since it’s one they were doing independently an outright cancellation seems like it’d be crazy. Unless they decided the whole ‘crossplay’ thing was unworkable.

  43. Jahkaivah says:


    If you take a look at the past articles, you will see that we love Love.

  44. toro says:

    What’s happening with Arkane?

  45. Dave L. says:

    A lot of for hire level design work, it looks like. They were working on a top-secret project for EA that may have been System Shock 3, but it fell through.

  46. eyemessiah says:

    Sentinel is tres awesome. Have you played Zenith?

    link to

  47. Babs says:

    Subversion seems like more than a hacking game to me. From the dev log it’s clear that he;’s attempting to construct all the moving elements of the world out of smaller components. ei a lift has an actual computer system driving it with motors etc.

    My guess is that the game will revolve in some way around ‘subverting’ the machinery of the city to achieve your goals, whatever they may be. Very exciting IMO.