[Music] and I'm gonna be talking not just about procedural generation but about procedural regeneration which is the act of changing level design in response to player action so hi I'm Heather I'm a cool person I like to think but one time last year I was at this bigshot industry party and this guy that I know walked by with a couple of CEO types and he just kind of turned to them and said that hey that's Heather she ate paper once and I haven't really lived that moment down quick content warnings for this talk there is one's live with miner zombie body horror and it I'll let you know when it's gonna happen I make games I make a lot of games these are some of the games that I made since the start of 2017 I made 30 games last year most of the ones that aren't up there are board games I like procedural generation I think it's a really cool topic and it I had to know it's very
near and dear to my heart so just a quick introduction to that because I want everybody to be on the same page for when this talk goes off the rails procedural generation is the act of using systems of rules to create large amounts of content now a lot of people when they think of procedural generation they think of games like Minecraft and spelunky where the thing that is procedurally generated is the level design itself the world of it if you will minecraft you by the way uses a purlin noise generation system while spelunky uses a series of rooms with mildly randomized elements within those rooms but procedurally but procedural generation can be used for a large number of different things so on the center right there we have a Ville Manik which is a game where you raise procedurally generated creatures with their own DNA and that DNA decides their habits and what they like and dislike and also just the shape of their bodies
and on the far right there we have a potassium edible max Kaminsky who I think is in the audience hi max which is a game about watching hundreds of civilizations die and you can't do anything to help them which is you know kind of a bummer but also it does a lot of really interesting things with world generation in history generation which is a cool thing so I really like procedural generation it has a lot of really cool things going for it and while it may not always make a lot of sense for a level designs perspective there is a definitive level design aspect in procedural generation and just it's important for everybody to have the same tools in their toolkit so it allows you to make large amounts of content in a fairly short amount of time it allows you to use systems based design which is my personal favorite kind of design I just think it's really fun and there's a great community involved with it on
Twitter the proctoring community is super helpful and super kind and just very willing to let you just submit to their scenes and it's a lot of fun to work with once it's up and running that is one of the main problems with procedural generation is that until you have everything you have nothing and that's a very difficult concept to wrap your head around at first so it is just a little bit scary but also it's a lot of fun and it's cool so what if we took that concept of procedural generation and looped it back in on itself this is the part of the talk where we go straight into mad scientist that mad science territory by the way and this is something that I have been calling procedural regeneration which is the act of changing level design dynamically in response to player action there are a lot of really interesting things going on here so I'm just going to take a second to pick apart that definition this is essentially saying you make a level and then you have an AI in control
of that level decide what the player can handle and from there decide which paths the player should take down that level and this is a concept that I've been calling the gentle guiding hand which allows you to use these less obvious systems to keep players on the right path and in general just Auto balance the game for them so that high level players encounter more difficult areas of the game etc and so we're going to be going through a couple case studies real fast we're going to be going for left4dead first followed by secret spaces followed by eternity left4dead is next one next slide is the one with the zombie body horror if that's not your thing look away of three two one okay so this is left for dead it is a all right we're off that slide now a four-player cooperative zombie shooter where you try to survive through hordes of zombies and make your way through the post apocalypse in all these different areas like a carnival and a mill and a
hospital and all these different places and as you go through the continent like there are a lot of interesting things about this game like it is a very well-designed Amba shooter but the thing that is interesting to me and relevant to this talk is that the AI in this game is all directed by something called the director which depending on how well you're doing can throw all of these different obstacles at you from spawning hordes of zombies to special zombies to deciding whether you deserve a health pack and to changing physical level design making players go down one path or the other or moving the choke points or just making it rain so that you can't shoot as well and it does all of this stuff to try and give everybody a TENS experience regardless of skill level which is interesting from an accessibility point as well the director does this through a combination of dynamic and Static procedural regeneration elements the dynamic ones
are things like placing zombies in the environment because that can happen in a spectrum of different locations and the director decides exactly for that situation where that zombie should go meanwhile static or scripted elements of procedural regeneration are the AI choosing between one of say two or three different scripts like if the player is above 75% health they go down this hallway and if they're below 75% health they go down this hallway and as a general rule dynamic is more difficult to implement for a lot of systems but also potentially more rewarding especially since there's this concept of a level that you can sort of know but you can never truly know no matter how many times you go down it that's a really powerful experience so let's just go down a quick example of procedural regeneration in a generic shooter game in action so that y'all can see an example of what this is all about so the
player starts out in that room down there with the star and they go north into this room which has three guards and the player viciously kills those guards but it takes a lot of damage in the process and so from here the game can decide whether to put the player into the easy path or the hard path and since the player took a lot of damage they go through the easy path and go to this room on the right with two other guards in it but if the player had done well in that previous room then they might go through this hallway in the north which is close corridors there are four enemies up there and then they have to still face through those two enemies on the rim on the right then both players regardless of how well they did go through this hallway on the right the player who needs a health pack at this point and this could be either the player who did well at first with a player who did poorly at first gets that health pack on the room on the left then both players would go up to that
room on top then follow that left in to the boss fight but if the player is doing well then that boss might that boss fight might have more enemies in it now I want to emphasize this is not two separate players this is the same player going through the same map with different realities these are different branching paths that the level could decide to put them on of course there are other examples for procedural regeneration as well which I'm about to get to now in the second case study and this is a game called secret spaces I made it it's really good I hope you all like it in it you grow plants to explore this gigantic living building searching for your girlfriend who fell down a hole because she's bad at not dying it's about a lot of things but for the purposes of our talk it's about the it's about forging a positive relationship with your environment and the way that it does this is through an
AI that I've been calling the tower AI as you move through the building the building forms an opinion on you it is an active participant in the story and also decides what level of difficulty you go against it rewards and punishes you based on your actions but this is less of a are you doing well at the game are you doing poorly at the game and more of a are you acting in accordance to the building's own morality system which is about like forming a relationship with the environment which moves plants around in the level which are the primary interaction points of the game and so if there are fewer plants in the level then you can't move as well you can't see as well you can't heal at all and yeah so it makes this decision both on the micro and the macro scale on the micro scale there's just constant day-to-day decisions like how many plants to the player have in their inventory but there are also larger scale events like how far has the player
progressed in the story and how have they reacted to specific rare events there's a lot of events in this game that only happened one every thousand playthroughs which i think is a really fun design space to work with so this originally started out as a deterrent from mechanic abuse and I'm not quite sure if you can see it but on the screenshot there on the right if you check the players inventory in the bottom center you can see that they have 100 of the rarest fruits in the game in their inventory which I figured should not be how any of this works that was sent to me early on in development this game was in early access for a while I got that screenshot from somebody and I said no no ok this isn't happening and added this AI into the game to try and stop that from happening so right now if you have more than 5 plants in your inventory at any given time there isn't a hard inventory cap but the game starts to heavily punish you by removing plants from the environment in order to give in order to get things from the
environment you have to give back to it in a sense and this ended up in addition to being a good way to balance the game also add it to the tension it made the game more fun because you had to make all these hard decisions and in addition it helped with the story and message in the long term so let's move on to our third case study now which is a game called eternity this one is still in progress it is a game that I am currently making right now as we speak in that laptop over there a very special laptop it is a stealth horror game about a monster who can change the past you play as a person who has entered this house in search of a cure for death and in that house what they find instead is the same reality rearranged over and over again they are a person shoved between hundreds and thousands of different timelines each of which is the same house rearranged and this is the extreme case
in procedural regeneration it's based on player action and movement like the others are but the world itself shuffles every couple seconds which turns what should just be a simple house into a labyrinth is this reasonable is this reasonable what we're just gonna have to find out the game still in development I don't have all the answers come to my talk next year I hope so there's only one monster in this game but because of the Warped nature of space this monster confronts you at almost every turn like you can evade the monster once and by hiding under a table and the monster will pass you by and then you go into the next room where the monster logically shouldn't be and yet there they are again because that next room swapped itself with the previous room so there is a quick distinction that I want to make between two main categories of procedural regeneration there is outward procedural regeneration which is the
kind of procedural regeneration used in left 4 dead and secret spaces in which the procedural regeneration moves outward it does it sees where the player is and decides where they deserve to go from there and the other example is inward proceed Jewelry generation which is buckwild and deeply experimental even more so than outward the primary example of it there is eternity now in order to display a like make throw into sharp relief the difference between these two kinds of procedural regeneration on the Left we had that example of outward procedural regeneration from before from before with the shooter and on the right we have inward procedural regeneration with a level in eternity that's a part of the floor plan of the house now it has a lot of the same major components going on like the player is still in there and there's still an enemy in there and the different doors in the environment can
change position based on whether the player is doing well or poorly in this case it's based less off of health and number of guards killed it's more based off of how many times the player has been spotted by the monster to try and achieve that sort of happy medium of course there is one major difference in that the level changes shape every 20 seconds that's the main difference between outward and inward procedural regeneration outward sees where you are and decides where you're going inward sees where you are and changes where you are so just real quick the way that I'm implementing this inward procedural regeneration for eternity is using a double buffer system which is something that is used often in graphics programming essentially the player interacts with the world as it is and a copy of that same world underneath it is constantly trying new forms new arrangements so that there isn't a huge
lag spike every time the world tries to rearrange itself because it's happening over time that allows for smoother gameplay even though it does mean a larger CPU load on the second a second basis it's not too big of a difficulty it's just something that I figured I should discuss how I implemented because this game is deeply weird and so an important thing to grasp in procedural regeneration is that there is no difference between a hard path and an easy path the hard path and the easy path are things in the environment but it's not a difficulty setting the player doesn't decide this I mean they could I'll get into that in a second but the level itself sees the player and decides what they're going to be comfortable dealing with now part of the reason this is interesting is because players are kind of stubborn they're not always the most willing to admit when they're not having a good time especially when something is too hard for them so having a system that can
just straight-up decide ok this is something that you would have fun with is something that could be deeply powerful as a design mechanic of course design and easy path could be a difficulty setting and procedural regeneration in addition to other ways of doing hard hardness and easiness by say making enemies harder to kill etc you could also just have the AI decide to be meaner to the player just have the players say ok I want you to give me your worst and see how that goes from there so let's go over the pros and cons of procedural regeneration real fast because no level design paradigm is perfect unfortunately though actually quick plug for my talk two years from now where I give the perfect level design paradigm it's kind of rule so the pros of procedural regeneration are that it is extremely cutting edge and you can do some really cool stuff with it and you will be innovating almost all the
time while doing this it's really good for game balancing on the fly like we talked earlier about the gentle guiding hand and about there not being any true difference between the hard path and the easy path the player can switch between them at any time depending on their performance in addition disorienting spaces are super spooky imagine a moment in a game where you walk through a door into a room you turn around and that door just isn't there anymore that's lightning in a bottle you can capture that with procedural regeneration and do other fun stuff with it in that same sense it has a lot of fun horror elements going for it because the game is always one step ahead of the player so the game literally knows what the player has done and conform itself around that fact for example of a player lingers too long in one room that same room might show up again in the same in a different context and in addition it works well with different playstyles not just in the sense of some players will be better at
the game and the game will adjust to that but also in the sense of say you are developing a shooter game where there are three regular weapons let's say a gun a shotgun and I don't know a sword and one super-powerful minigun now you could and now the minigun is a lot of fun to play with because it just shreds through enemies and you really don't want to nerf it but also the player can access that thing at any time how do you deal with that difference in weapons strength well a normal game would say that you need to just nerf the minigun but you don't have to with procedural regeneration because you can just because the AI will detect that the player is just tearing through hordes of enemies and will just throw more and more enemies until it's reached the same level of difficulty which is something really interesting because it means that the player is still having a good time and fulfilling a completely different kind of power fantasy than they would have otherwise of course there are some
cons of procedural regeneration and right now it is a very specific design space right now it's mainly used for balancing and horror purposes I do want to see this changed over time I'll talk about that more in a little bit in addition it is a relatively unexplored field there are very few games that have done things like this if you have any more examples after the talk I would love to hear them to be honest in addition disorienting spaces are super cool but they're also super confusing especially if you don't want to give your player a map that's something that I have experienced with play tests of eternity is that the player gets turned around so easily it's very difficult for them to know where they are where they're going or even what the point of the game is and in addition procedural regeneration is difficult to program I use the catchphrase better out than in here that's a reference to on the difference between outward versus inward procedural regeneration outward procedural regeneration is easier to program it makes more sense from a level
design perspective and it also makes more sense for a player perspective and while in word procedural regeneration is super cool it is also super unexplored in addition procedural regeneration can be risky and triple a development that's something that's true of all procedural generation paradigms is that it is and there's that same concept of you don't have anything until you have everything you can't really start playtesting until you have the world and you can't have the world until you've done hours and hours of coding so that is a minor problem with a procedural regeneration and in addition Kate Compton has this concept of procedural oatmeal and procedural regeneration error in procedural generation which is that you can make a bowl full of oats and each oat has its own geometry and its own history and all of these cool things and you could spend years working on this bowl of oats what are the players gonna see is a bowl of oatmeal they're gonna eat in three
seconds then it's done so you need to make sure that you have enough variation in your procedural regeneration formula so that players will have that players will actually be affected by the differences that you're making in the game it's code for example if you had the player if you had the AI choose for the player to either go down a hallway with four enemies or a hallway with four enemies that's not going to affect a player you need to actually make there be a palatable difference in order for that system to properly work so in his procedural regeneration right for you I will admit it is a relatively unexplored design space as I have said but it's worth checking out this is another tool that you could put into your tool kit and in a game maybe five or ten years down the line you might find a really cool use for this and no level design paradigm is open Asia which is what makes level design so wonderful if we all just did the same thing every time it would just be boring I do want to
imagine a future with you for a second real fast where every game ten years down the line uses these wildly different forms of procedural regeneration we're all levels are in some way interacting with their players where every level is its own Galatea we are not bound by the shackles of reality and we as level designers can do some really fascinating things with this concept and I am deeply excited to see where y'all take it so thank you my name is Heather Robertson there's my website and my email and my Twitter we have a couple minutes for questions if people want to step up to the mics oh yeah the microphones are there and there if anybody has any questions
[Music] hey Roy sir talk so good I don't know just kind of did it that looks like we got one hi hi so how would you prevent players from in this case of frosted procedural regeneration how would you prevent players from for example abusing the system in terms like going on the hard path may be like taking damage on purpose and then just going the easy way and then taking damage again going the easy way that definitely is a thing to consider I could see this really becoming a problem because you can no longer easily define how a speedrun works because I mean like you probably you would have to figure out exactly what is the correct path to follow and how to take damage during certain parts which then again is it's interesting design problem in and of itself but anyways to answer your question I feel like players don't want to
intentionally take damage they will they will just keep doing their best until their best isn't good enough and they'll be put onto the easy path and it doesn't just have to be hard and easy there could be any number of paths going as a spectrum between doing really good and doing really bad and if they decide to go down the path of having an easy time for a little bit I feel like that's their decision to make alright thank you thank you looks like we have another couple folks so I've never played left 4 dead so this might be something that a bunch of people in the audience already know the answer to but you know I think as any game developer knows players are often very quick to encounter something they didn't expect and just decide it's a bug and so I'm wondering with procedural regeneration when you are you know on this playthrough they're doing really well so you send them down the hard path and on this playthrough they're not doing so well so you send them down the easy path do you have any tips for how to sort of indicate to the
player that there's a reason why this door worked last time and it doesn't this time hmm that is an interesting conundrum I feel like a a good way to potentially do that is to hide the existence of both doors until it's decided to go down one path or another you could do this fairly easily let me go back to this example here on the Left where instead of having the two doors in the same room you could have them be down a hallway and be obscured by line-of-sight that way and I feel like that would obscure the problem for the player another way of potentially doing it would just to be frank with the player and say this is a thing that we're doing the game will auto balance itself to you and I feel like some players would really be into that thank you hi thanks for the talk really enjoyed it regarding play testing how difficult was it when you were doing a lot of this procedural regeneration and kind of locking a scenario through a play test because it seems like every time you
would start to play tests over the data would be different and so I like explained maybe some of the challenges you went through with play testing thank you for the question umm I am still like very early on development for eternity so I can't super answer questions for playtesting that really well but I can talk about secret spaces really fast which uses that outward procedural regeneration element and in that game the play testing was less about the longs the large-scale game it was always about the moment-to-moment and that moment-to-moment is the same the same regardless of the overall structure at the level you can see oh just one minute left and do you want to ask your question real fast and that'll probably be the last one unless I'm really fast so in the case of eternity you said that the player was going in search for the cure for death yes are we assuming that the cure for death is in one of these
procedurally regenerated rooms then okay so this is going to get a little bit spoilery but like I feel comfortable doing this in this room in eternity the player is searching for a way to stop time and the monster in the house controls time so it's not a MacGuffin then basically so it's not just an object it's not an object it is a creature I was going to ask if they were capable of just finding it walking through the front door and saying oh it's there on the nightstand alright that would be cool that's sort of like the concept in Fallout 2 I believe where you can just kind of go to the final boss immediately if you just head west yeah all right um do we have time for one more question no we do not okay cool thank you all for coming [Applause]
Auto-generated captions from YouTube — names and numbers may be misheard.
Distilling the key insights…
Game Developers Conference · 2026
Jason Schreier · 2026
Comments