I talk about my thoughts on how games can use both procedural and handmade content, based on how I have used it in my past games and how I have wanted to use it in those games.
I'll never get over the fact that WB patented it and has done absolutely nothing with it since. A horrible decision that is going to completely halt any innovation in one of the best new procgen systems for decades.
its implementation leaves much to be desired in both games it’s featured in, and we’re never getting another one so it’s essentially a dead mechanic now in terms of innovation potential. Of course, on the indie side, mcdickie games were using it before and are still using it now. And even though I really love wrestling empire, those games are more akin to a really good shitpost in terms of production value.
@@ElvenSonic "essentially a dead mechanic now" IIRC there's word that WB's in-development Wonder Woman game builds upon Shadow of War's nemesis system. But also what I've heard from people who look at the legal side of industry stuff is that the patent is somewhat vague. Maybe, for all we know, that could mean a deeper purpose of the patent is more about deterring other studios via the risk of *timesink* from legal proceedings, rather than WB assuming they've got actual assured court protection from anyone else being allowed to build upon those mechanics. Could help explain why one of the only "big" games to include something approximating the nemesis system was Assassin's Creed Odyssey, since Ubisoft is a powerful enough juggernaut that they probably already have in-house lawyers ready to go.
There is merit to something being finite. It can be completed, it can be exploited/predicted, you can discuss it with other people. That feeling you feel during New Vegas, the melancholic feeling of this is it... it's a powerful and addictive feeling. So many games/movies/series nowadays want to go on forever... but finishing of something and finishing it well is something the audience secretly craves and the investor laments.
A good ending can really seal the deal on something that was already awesome. And so few story writers get it right or even get the chance to write it. Case in point - True Demon Ending for SMT: Nocturne. What an amazing game ending, you ascending and becoming a champion of the mightiest being in the game setting. This was then made canon and your main character reappears throughout the series as final extra boss fight to clear, if you're masochistic enough to try.
I wonder how much experience Tim has with games that rely heavily on procedurally generated content, like roguelites and Dwarf Fortress and such. In my experience procedurally generated locations and characters get stale really fast, once you've seen half a dozen permutations it stops being interesting. On the other hand procedural generation that affects the initial faction dynamics does make playthroughs feel more fresh, like say in Mount & Blade some faction starts snowballing or in Dwarf Fortress your biome and neighbours affect your game a lot. So the smaller scale stuff is a lot better when hand crafted like characters, items, exploreable locations, but the larger scale things like map level biomes and faction power dynamics do benefit a lot from randomization.
I think part of this is just that the majority of designers doesn't really think about the generation too much. Bashing noise functions together to generate landscapes and dungeons is a very surface-level implementation of procedural generation at the end of the day. A true high-concept randomized game would have building blocks designed from the ground up to lean into generation, so that you can get those highlight rooms and encounters and everything doesn't feel the same. Of course, letting your generation have peaks like that runs the increasing chance of generating imbalanced or unsolvable runs, so there is a lot of consideration going into this beyond turning the RNG knobs up. But imo, most current procedural games play it very safe and use the technology just to provide comfortable stream of content instead of being interested in pushing the possibilities.
Ironically, Dwarf Fortress generates the smaller details, but it's larger structures, like civilization types and different groups of semi-random monsters, are almost like premade content. Civilizations are a lot more elaborate than procedural content usually is, but, for instance, goblin civilization has the same structure, same with every dwarf or elf civilization, and then they're rearranged in different positions in each world. Meanwhile individual dwarves or places on the map have a lot more flexibility in their generation.
Exactly what I was feeling. I have a lot of playtime in Minecraft which is built entirely on procedural content, and it has sustained an active player base for a really long time. But even with this "infinite" content, the developers still (rightly in my opinion) feel the need to keep adding more features and content. It seems like procedural content can take just as much effort to generate meaningful playtime as hand crafted, at which point it comes down to player preference. I usually start to notice the patterns in procedural content really fast and it lessens the experience for me, while others are perfectly fine to play them for thousands of hours.
The real difficult thing in procedural content is making it varied enough to remain interesting for a long time. Humans are great at noticing patterns so we're really hard to keep immersed with content generated through a rigid set of rules as those rules start shining through brighter and brighter with each time you play through a piece generated with them.
I think my main issue with a lot of the games I have played that rely on procedural generation is the eventual feeling that I get that it doesn't actually matter where I go or what I do. If the proceedural generation is as likely to generate a quest/mission if I go North as it as if I go South, it doesn't matter which way I go. If I know that any given cave or dungeon was generated with a roughly fair difficulty based on my level, and will give me a randomly selected treasure appropriate for me, then it doesn't matter which one I go to. If there is no real central quest or mission, then exploring the world and gaining power really has to be fun in its own right, and while sometimes I can get into a groove with a game like that, more often I find myself asking why I bother.
That's a good way to put it, and I think is the key issue with Starfield. Without systems in place to force me to live with my decisions, nothing matters.
A lot of proc gen is just sterile, the distinction between going north and south doesn't have to be so meaningless. I think this is an area of games we're yet to see the best of, more so than most other aspects.
@@Gnurklesquimp2 Agreed, I think there is a lot left to be attempted involving more robust procedural generation but it requires an attention to detail, worldbuilding and a development plan that incorporates procgen as a cornerstone from the beginning that many developers are going to be wary of for fear of not releasing a successful game. There are very few games out there with handcrafted content that take things like the emergent impacts of physical geography and socioeconomic realities into serious consideration when designing their worlds (one of my big problems with bethesda games even though I love them, they feel like disney theme parks more than living worlds) let alone developers designing robust procgen systems around such things.
@@sieda666 That last thing you covered is something I often consider. AI doesn't have to be lifeless, in fact, it can bring life to complex things we struggle to by hand. This would already help with mostly static worlds, but I think we're gonna see worlds that live and breath through ecosystems in every major way, something I also wished for with every TES game I played. Hell, some of the budget paying for people who'd otherwise be making things by hand can go to experts that merely review the AI's work. Not as simple if it generates stuff on the spot in playthroughs, but otherwise doable if you give them some tools to make it faster.
My biggest gripe is particularly focused on the latter half of your comment. If a game has really good gameplay, I couldn't care less if the dungeons I'm running are randomly generated and predictable, because it's the gameplay that's keeping me hooked. But that's rarely the case with games that rely on procedural generation. I loved snap map in doom 2016 because it meant I could play the game more, and in new ways. But it wasn't procedural. If it had randomly generated levels that went on forever, I wouldn't care. The gameplay is just that good. Fallout 4, on the other hand? Its gameplay isn't bad, but it isn't good enough to justify playing for the sake of playing, in my opinion, and I would avoid the radiant quests whenever possible. My point being games that have procedural generation rarely have the gameplay to back it up, and gameplay that is just that good that you'd wanna play forever always seems to be in games that are seemingly over wayyy too quickly. I think binding of Isaac and ziggurat are kinda good examples of procedural content done right. But those aren't in the same wheelhouse as something like fallout or skyrim, the random generation is literally part of the gameplay, it's not there to extend the lifespan of the gameplay.
As others have mentioned I feel like the type of procgen being described by Tim is being pioneered by rogue likes. Dwarf Fortress, Cataclysm: DDA, and Caves of Qud come to mind
I like the Dwarf Fortress world generator. I prefer generators where the final product is a result of a sequence of rules based inpainting, with the setting sliders making use of the noise, instead of a simple fitting together of prefab tiles, and DF using volcanism and rainfall erosion to generate mountainranges, which in turn will inform civilizations is very appealing.
Wayward Realms made by the guys behind Daggerfall and a few other people is going to have procedural generation that accomplishes crazy things like what you’re talking about. They’re planning on making it into this interconnected world with all kinds of different things constantly changing how everything interacts. It’s pretty crazy what they’re planning. There will also be a bunch of handcrafted stuff to ensure the game still has a soul.
In S.T.A.L.K.E.R Clear Sky (one of the least liked Stalker games, which I think is unfair), it had this _really really cool_ faction wars system where NPC squads would wander around and interact with one another. Bases would get attacked and taken over or maybe the attack was repelled, but every interaction was signalled on your PDA so you could head over there and help out one side or the other, or you could just wait until it was all over, pick off the survivors and loot everything It was amazing but so few games use that system There was a mod for Kenshi which did something similar and I crave that in open world RPGs my brother and I also use the term forever game to describe theoretical games with worlds that are as close to living as we can imagine :P
If you like Clear Sky which I liked too, probably the best proc-gen, open play STALKER mod right now is GAMMA, look it up on youtube, it’s based on Anomaly but with a lot of annoying ideas stripped out and core gameplay loop enhanced. You can work with any of the factions from the Free Stalkers, through Duty, Freedom, Mercs, Ecologists, etc, with rewards that tie into the factions. In a lot of ways it plays like Mount & Blade: STALKER, and I love it! Edit: Also unique gameplay modes like Azazel Run, where if you die you become a random stalker in the zone of any faction with whatever gear they have, faction relations. It also still has the base Anomaly quest line if you want to do a main story too.
GTA san andreas had turf wars between gangs that happened periodically /dynamically, and you could choose to participate, added quite a bit of life to the city.
i guess this is just a trope of the genre to an extent but i'm a bit put off by the idea that the whole point of this is to just make a game as big as possible. as i near my 30s i'm becoming increasingly aware of the reality that my time is most certainly NOT finite and i don't like the idea at all of just filling all that time with just one piece of art. there are so many incredible video games (and movies and shows and music and books) that i want to get to experience in my lifetime and it makes me pretty anxious to think that many of them are RPGs that will take several hundred hours each. i understand your enthusiasm tim and i'll admit there are no doubt a solid handful of games that i love enough to be able to play perpetually if i was forced to, but i also have the choice of playing many many more wonderful games instead and broadening my artistic horizons and i'd much rather do that
I came to really enjoy Daggerfall for the scale, it really felt like my own story because of how lost I got among the many cities and towns I based out of. I tend to replay handmade games more, but I am looking forward to the Wayword Realm's upcoming title that explores these large scale worlds. In terms of quest, story, and even NPCs being generated by Ai; I'd have to say that it should be used to assist the existing experience. The main story, main hubs, and characters should be hand made. Games are becoming like books in many ways and should reflect intent beyond 'making money'. So they should have a crafted experience.
Wow, didn't expect people to be talking about Wayward Realms here. It feels like nobody knows about them and their game (I only found it, I think, by googling Daggerfall's original designers and what they were up to). I also liked Daggerfall, trying it for the first time a couple years back, but what absolutely destroyed my desire to play is the god-awful dungeons. They are a perfect example of procedural content gone wrong, in my opinion. Just soulless mazes of the same corridors and rooms, randomly rearranged over and over...
@@BuzzKirill3D I can't disagree with the procedural generation of the dungeons being not great. But when you consider the time for it it was fantastic. Thank god for mark/recall though. Daggerfall was almost 30 years ago and to me is still a cornerstone of procedural generation for areas. Honestly with Oblivion and Skyrim (and some of Fallout 3 and 4) I g to a little tired of the "dungeons" almost always having a shortcut back out to streamline the handmade dungeon experience. The sense of exploration that those procedural dungeons had was great, you could get just utterly lost in them for hours.
You were saying "imagine" when talking about procedural areas and quests, but there's already a game that does that. Dwarf Fortress is a great example of both the strengths and weaknesses of proc-gen: it's infinitely replayable for the same reasons you mentioned and more, but everything in it lacks the detail of a more handcrafted game. It's also been in development for an exceptionally long time, although a lot of that was caused more by feature creep than issues with the engine.
The most interesting recent exemple I've seen of truly good procedural generation I've seen is "Shadow of Doubt". It's a detective game with procedurally generated crimes, in a procedurally generated city district, with hundreds of generated citizens. And it adds a very important ingredient to the procedural generation : *simulation*. It simulates what everyone is doing, they go to their jobs, leave notes etc... It is very finite in what it represents: the city is small. But it is very dense, and it feels very coherent. Because the rules that place items in the world make a lot of sense. I like No Man's Sky, but everything in the world is very temporary, which makes it harder to care about it. The different elements lack connections, two individuals will never communicate in any meaningful way.
Shadow of Doubt is still in EA, right? Personally I hope someone makes remake of first two Fallouts filling it to the brim with procgen events/things which would however make complete sense....like Tim said. Although my realistic self thinks that at least for first few years of procgen boom the games working on that mechanics/algorithms are going to be bad and boring...with few exceptions though......since the procgen stuff will be only as good as the core mechanics/art/story created handmade. As long as we will have AAA studio releasing bland and shallow games then not even procgen can change it into masterpiece.
All your examples sound decent when just initially brainstorming, but making this all work while being fluid, making sense, and not being obviously prefabricated cookie-cutter looking assets is really tough. The explosion of procedurally generated games in the past decade have been interesting, but all of them that I have played, I instantly wish they just hand-crafted the entire thing. The single game that I have played where it works is Binding of Issac, but even then, that is a relatively basic system that just happens to work based on the strength of the core gameplay loop, as opposed to the complexity of, say, a randomly generated Metroidvania or open world RPG needing to nail exploration and proper item progression.
Also, the idea with procedurally generated quests is just one quest you can do 50 times. These are not 50 new quests, not 50 new stories, thats 1 quest you do 50 times in different areas. Its gets really old, really fast in my opinion.
Especially when very few variables change, like in the Bethesda style radiant quests, where two names and a location pointer is the only thing different between iterations.
Hey.. glad you’re here.. we lost our friend in the cave 400m away and need you to go pick him up. Hey wow.. so glad you’re here.. my group got stranded can you walk me 1500m to my starship? This is like half of the random encounters so far in 60 hours of SF
The point is it could be so much *more* as it's built up and built out. Theoretically there's no reason it couldn't generate entirely new types of missions, generate actual stories with plots that take you to specific places for a reason rather than just randomly putting a McGuffin somewhere. It wouldn't be easy, but I think there's an entire ocean of untapped potential because devs stop at "welp, it technically works." and anything beyond that requires a lot of work, R&D and money.
@@Xbob42 " generate actual stories with plots " -- I'm with him --> "'Each time, the team collectively believed - believed down at their CORE - that this was possible,' [David] Gaider said.... 'Just within reach. And each time we discovered that, even when the procedural lines were written by human hands, the end result once they were assembled was… lackluster. Soulless.' [David] Gaider believes developers - indie and AAA - and executives will still continue to push for AI in games, but that it will only result in the same conclusion BioWare drew and end up being a colossal waste of time and resources." And perhaps even put off enough RPG players to make them finally sit up and take notice. Of course, that won't happen until and unless their sales drop off even more sharply than they already have.
In a lot of ways I usually look at it the way I look at Braid etc., they think of a mechanic and see what can be done with it through a process of imaginative exploration, the best mechanics they find get turned into incredibly elegant games.
I watched a talk about dwarf fortress I think was a gdc talk The cats were dying and they creator had no idea why but apperently they were walking in bars and then cleaning their paws and then dying from alcohol poisoning. th-cam.com/video/6yWf6BHqiWM/w-d-xo.html
I appreciate your enthusiasm on the topic, but for me procedurally generated content is just inherently uninteresting. I would just much rather experience a story that some real person has put thought and emotion into. The only example where I can appreciate proc generation are rogue likes. But even there I have grown tired of the absolute sea of roguelikes out there..
There's plenty of adventures built on the premise of procedural uniqueness being at the core of the gameplay. Just because AAA hasn't done an amazing job of it lately doesn't mean it's inherently bad
6:50 While mentioning fallout, I always wanted a 3D fallout where you could travel long distances like daggerfall. Always felt a 3x3 km2 Map was a bit small. You keep smaller important areas hand-crafted and the areas between major locations procedurally generated. Your idea is basically what I always wanted in a game.
Proc-gen is one of my favourite features in games. Great to hear more of your thoughts on the topic! I love roguelikes and I'm working on my own using wave function collapse.
Thank you for your videos! This one in particular was a great discussion. Totally agree on the value add of procedural generation and the potential to add richness and narrative.
The problem with Starfield, from my understanding is although radiant procgen caves and buildings very much do appear on planets, they often repeat a bit too often because Bethesda hasn't got enough handmade content to make it feel unique. Examples are caves with the exact same skeletons in, or outposts with the exact same layout and NPCs in the same place. I also personally think the 1000 planets is a bit much at this time and they perhaps should have reduced the scope a bit to allow for more unique procgen radiant encounters.
It's probably impossible to make enough handmade content to make things not repeat. Think about it. If in 5 hours you encountered the same location 4 times, then even if they doubled their handmade content you'd still come across it 2 times in 5 hours. And even if they quadrupled their handmade content you'd still see a repeat every 5 hours. The truth is, you can't make enough handmade content to solve this problem on this scale. What they need is AI generated content but AI isn't going to produce Bethesda level content to rival their handmade stuff. And keep in mind that the more "radiant" a game becomes the more bugs can be introduced out from under the supervision of the team.
I don't like the idea of “forever games”. I want the game to eventually come to some conclusion. Sandbox is cool, but I really doubt it's possible to scale quality of random generation to the level of content made by a creative person. That said, I don't mind a bit of it if it's a well organized system that takes a large set of parameters and turns them into small quests. It's always nice to have some side activity that isn't obviously staged (looking at you, cp77 🤨) and doesn't repeat several times in a row.
I mean you can have both, there is no reason a game can't have a campaign with a definite, satisfying conclusion and also infinite side content. A lot of the games that people really play for years on end though have a very small amount of content, but people don't get bored because the mechanics on their own are fun to enough engage with and improve at endlessly (e.g. fighting games, chess, tetris.)
Tim did you ever play dwarf fortress? (Or something similar like rimworld/prison architect). I'd be curious what your thoughts on them. Since i knew fallout and dwarf fortress i always wanted to make a game that combines them both... but then kenshi and rimworld came out so they kinda did that... (Kenshi is awesome too, btw)
You'd love Elite Dangerous :) It's a forever game with the entire galaxy to explore. It could use a bit more "handmade" or better procedurally generated story content, but it has a lot of the exploration features you were mentioning :)
I think some of the best procedural/infinitly-growing feature in a game was the nemesis system from Middle Earth, especially when you brought up new stronger people showing up in the game world it immediately made me remember how often the difficulty changed in Shadow Of Mordor based off of the way I fought my current enemies
Hi Tim! You're my hero, these videos mean a lot to me and make me smile every time I see them. They're also very damn useful for any game dev to check out. So, with that in mind, i'll drop in a question pertaining to my occasional struggles in development. When stuck in a development rut, or any other applicable synonym of writer's block, what methods do you believe are most effective in remedying this issue? How do you avoid getting into a rut? Thanks so much for all of your videos and help, as well as for taking the time to read our questions and such. You're amazing.
Some of these have been implemented, in more minor manners, in other games. Like for example, when you describe wanting the world to feel alive, I remember black and white having an almost "ant colony" vibe to it, where you play as a god, you can direct your worshipers to certain points, and you get to see them essentially do it in their own way. I remember settlers 3 did something like this too. The idea about a "special" NPC arising at some point to challenge you, that was actually closely implemented by the shadow of mordor/war games, the nemesis system. Essentially when you'd die to a random low level enemy (even by accident), he would get promoted and in the game rise ranks, he'd become a commander, eventually if you kept dying to him, he'd become a warlord and keep increasing in power, both physical and in terms of the armies he could command.
thank you Tim, i was really reluctant about procedural content until i saw this video, you made some good points and i'm actually hopeful for a few games that'll come out soon, great video as always.
If I understand you correctly, you are talking about a living world which evolves within your setting for a player to immerse themselves in. I love that idea. For me, that’s the holy grail for an mmo rpg. An mmo rpg where the world does not revolve around the players, but the players nudge the world in a direction.
Tim, I love your take on procedural content! We've barely scratched the surface of its potential for fascinating gameplay experiences. I hope to help remedy that as I use it extensively in the action-RPG I'm (slowly but surely) developing. 😄 Anyway, thanks again for sharing your thoughts!
Tim, I just want to say that thank you for using this platform to talk about game design. I have been formally trained in graphic design and the fine arts but recently I have been making my own game. Your philosophy or even just what is made into a game has really opened my eye to what even it takes to make a game. I just want to make sure you know that you really have made an impact in my design journey. Thank you. Also are you working on anything currently?
I think the epitome of these being used together in harmony, is Ultima Online. That's one hell of a game; static cities, dynamic in-world player housing. Usual random spawns, able to tame any entity in the game and it becomes unique.
Arena Dagger fall comes to mind. Imagine a game with core content, but also has stuff like a dnd hex crawl, with random generation , random interactions. merge the rogue-like with the RPG on a complex and never repeating scale.
Diablo 2 had procedurally generated maps that work well and that game came out a long time ago. Procedural content can and has worked. Developers have also experimented with it unsuccessfully in the past but I think all that effort is worth it to realise its potential.
It really depends on how it well it's implemented, some of the best examples I have tried are Minecraft, Sid Meier's Civilization, and Mount and Blade. Some of the worst would be Bethesda's radiant quests, but maybe the simplicity of the gameplay is to be blamed there. If stealth in Skyrim had the depth of Metal Gear Solid, those Thieves Guild radiant quests might be fun.
The only game I have seen that has procedural generation that is rich enough to create interesting stories is Dwarf Fortress. I could tell many stories of silly things that happened during my games or tales of how my fortresses ended or a story about the one time I tried to pump water and accidentally made a water slide of death.
So, this makes me think of Dwarf Fortress, which I see others have also mentioned in the comments. To me, procedural content can lead to sludge or to diamonds. In DF, it leads to diamonds. In lots of other games it leads to sludge. Why? Well, DF is set up as a history generator. The game generates a world with a mythology and a world history with detailed events transpiring between settlements and individuals. This means that what the game presents you with is not just procedural sludge, but people and monsters and places that have an evolved history and connection to the world that they and you are in. Things have meaning and a history. It's not just perl-noise in the shape of game assets or whatever. And ultimately, to me, a really big thing that I play games for is history. Sometimes, it's the histories that the game's creators have decided to tell; and sometimes, like in DF it's the histories that evolve.
Great insight into the potential of procedural content. I really enjoy well-done procedural content and believe it’s the future of RPG’s with sim elements like Daggerfall.
I recommend everyone here try the Fallout Fixt mod, so you can see a restored version of the original super mutant invasions. It has a similar dynamic feel to what is discussed in the video.
Thank you for the great insights Tim! I love the application of procedural generation when done mindfully and when being appropriately explorative so as to avoid excessive repetition! I think Arcanum's dialogue was successful in that. Can you make a video on the procedural perk generation you mentioned? That would be interesting 😊 Have a great day!!
I cannot express how surreal it was watching this video and you explaining the same things I've been so interested in over the last couple of years. I've always wanted to create a "forever-game" version of Fallout, and procedural generation is so interesting to me that I'm currently shaping my senior project for undergrad around making a general-purpose procedural tilemap generator for Unity that will (hopefully) be able to include structures and dungeons and NPCs. I'm bookmarking this specific video for future reference if I ever have trouble explaining what I want to create to other people, and will be referencing it as a source of inspiration (even though it came out after I started my project). Thanks for the great insight as always Tim.
I literally had similar thoughts about procedural generation, about what is possible. I plan to create very sophisticated algorithms for my space exploration and combat game, factions with alliances and feuds. I also thought about how to avoid making the generation too stale or boring. Because that is the core problem. I too believe that doing the algorithms properly is the solution to the problem. Doing bad algorithms is like doing a bad job in "handcrafted" elements. I think it's just more challenging to make algorithms right, as they are quite heavy on math and logic. One example of how not to do it: Assume planets can have attribute A, B, C, each of which make them interesting/relevant. Without them, they are useless to the player. Giving a chance of 10% for each of those to occur means the chances for 2 of 3 to occur is 1%, and 3 of 3 to occur is 0.1%. But we don't want the player to comb through thousands of planets to find one that is interesting. Solution: Either increase the rate of multiple of A, B, C to occur higher, or create mechanic that allows the player to quickly eliminate irrelevant planets or to find relevant ones.
Honestly, procedural generation in SF detracts more than it adds. The main points of Bethesda games were exploration of handcrafted universes and complete immersion. First is torpedoed by procedural generation, second by constant fast travel and cutscenes.
I would argue that's Tim's entire point: Bad procedural generation is the issue, not procedural generation overall. Also the handcrafted engine Bethesda used was so fucked up they couldn't do seamless travel when like 12 nerds in the UK did it years ago for their indie space game. The fuck is Bethesda even doing?
@@Xbob42Because rebuilding their engine from scratch (which is used for more than ONLY SF) just for seamless flight into planets is not worth the huge investment for a game that isn’t a space sim. NMS’s bespoke engine was specifically created for a single space sim game.
@@tronam Yeah I know, they made Starfield shittier because they didn't want to put in work or use an engine better suited to the game they made. You can say it's not a space sim all you want but good luck finding anyone fuckin' dumb enough to think the game is better filled with nonstop load screens and fast travel. Classic Bethesda.
You are right. People in general do not understand the potential of proc gen, but rather focus on the limitations of past games that tried sell it in their marketing...
None of the procedural stuff works well when it comes to narative design and dialog. That's why in Minecraft, there is no dialog or quests. As soon as you introduce the narative aspect it all becomes boring, repetative, fake and stitches are revealed. So no, it's not a good idea to lean heavely on the procedural stuff in future Fallout games. Starfield is a disapointment in large because there are very few quests.
Procedural generation is something I've wanted to explore further, almost enough that I may have pursued a PhD around it. (I wanted to do a multi-node parallel processing cluster to run a Dwarf Fortress and greater level of world sim that constantly runs.) The problem you touched on with current examples of procedural generation... They simply don't go far enough. A lot of the procedural generation systems I've seen out there are limited for one reason or another, and they simple don't go past the bar they need to. And a lot of examples front-load on initial generation... but do nothing later, which makes it easy enough for players to even subconsciously detect the pattern. I think current gen hardware finally has plenty of processing power and storage capacity one average to really do a procedural generation properly. I can imagine layers of small scope generation systems being layered upon each other and the output laminated to a "base" state. Then, you take this laminated state as input another set of generation systems to layer on their own laminated base state. And you repeat for each category of meta-layer: Geology, Biology, Anthropology, Technology, Sociology, Political, and Psychology... So, you end up with a bunch of laminated based states that are like transparencies upon each other to project a new world to the player. Now, the crucial thing is the ability to call upon the systems you used prior in the game to reprocess any deltas written upon the base states to generate new base laminate states and cascade the change up the stack. This feedback loop is what's missing in a lot of procedural generation systems out there. This is why stuff feels so stale, because it really doesn't change to what you do. If the player hunts a particular monster, it should start showing up less and less, until it's either extinct or moves to another part of the world where it doesn't feel threatened. If the player stirs up drama between two towns, I expect to see patrols around their borders and maybe even checkpoints on the roads. If I'm the personal grim reaper for a particular gang, they should either be sending hit squads after me or clear out when I show up. For procedural generation to work, the world needs to react to the same degree it was generated. For anyone that's played a tabletop RPG, the best campaigns have always been when the world reacts and changes to what your characters do. It doesn't matter how it started, it matters how it changes. Procedural generation can allow for infinity permutations of change depending on the player, it's just not utilized properly.
To add, it has to change not only to people, but to environments as well. People say the problem with Starfield is everything is random, therefore nothing feels unique or immersive. The truth tho is that nothing feels cohesive or organic. When you get to a world with the pirate captain, you see a military base that was a secret lab to an opposing faction. You hear a monster got loose. Raiders have control of this world and are freely traveling amongst it. This tells a big story that the world needs to build around. Maybe the raiders have repurposed machines to the factions left over items to their needs. Maybe they made this beast its pet out there? This could be a theme that the procedural world builds around to make it feel alive. What I explained tho, this location is part of the story. Now you go to a random planet. It just has random stuff happening for no reason. Nothing feels immersive because the world isnt reacting to itself or the content it made for itself. Imagine, that the world created a flower...a poisous flower...now its draining the nearb6y lands of water and creating toxic pools. The trees begin to be all deformed due to lack of nutrients. The people here now survive by living high up to escape the fumes. They harness the nature to create weapons since the outside world holds food that they need, that or use it for drugs to create a crime lord planet. This displays reasoning and logic with itself to create random worlds and experiences that react to the algorithm. This is my biggest disappointment with these types of games. They half ass it. They forget they have to actually work to make it feel alive. Actually have to think bout how things interact.
A lot of your points in the second half I would like to agree with because in theory they're great ideas but after Starfield I'm becoming more and more against it. I've already given up on Starfield purely because I'd keep running into procedurally generated environments, quests and NPCs and it was all just so bland and obvious. It also ends up unintentionally drowning out the hand-crafted elements of the game and pushing them aside. For almost every marked/pre-made environment in Starfield I kept wondering if it was procedurally generated or not and I mean that in a bad way (as in not a compliment to Starfield's procedural generation mechanics). In my opinion writing and worldbuilding are the two core pillars of RPGs and having random generation take too much of the load off can ruin the whole experience. I have no interest in completing a revolving pool of generated quest types no matter how good the gameplay is. I don't think there's a way for a game to have that ideal amount of unlimited content and still be good. I'd much prefer games end and leave me wanting more than take shortcuts to try to give me everything and ruin the experience
The procedural generation does not always need to run on the players computer. It can also be used as a tool to pop out a lot more content during development. Mixing algorithmic generation, using handmade elements and doing a manual pass over the result / manually select the best generations.
12:38 I remember being absolutely blown away when I was playing Anarchy Online for a bit with how there was a literal floating city that would appear at random and my non paying newb self didn't understand how to get up there. It was awe inspiring. And this may be a bit of a stretch to say this is the reverse of what you were discussing but that game had a novel series attached to it and somehow players' actions would sometimes be a factor for what would get written and focused on
*The problem is that infinite in this context is still just finite with a little mask on it.* Any content that you design an algorithm to make is in effect really still just you making content. No matter what you do you can never design an algorithm to becreative because there's no intelligence behind it beyond what you are putting into it. And even if it was done the best I possibly could and you somehow created a system that could endlessly generate conflicts for some kind of reason for an endless numbers of places, let's say by running 100000 games of dwarf fortress in the background to create a somewhat interesting and seemingly unique history for anywhere you go, you're still going to be confined by the parameters that make up the algorithm. Let's say you even set up a large language model for each character in each one of these places so that they're all populated by unique NPCs with a history that can talk with unique dialogue in a infinite number of voices, so you have a infinite number of infinite places all with unique histories and characters that interact with one another all according to an algorithm. What's going to happen is that even if then somehow ends up being compelling at the start, it's all going to get incredibly see me with the same kind of characters having the same kind of conflicts that all talk in the same kind of way no matter where you go. So instead of just having one crafted game with a finite amount of content, now you have a averaged out smooth version of that game but stretched out over a hypothetical infinity. *That sucks.* For as interesting as your algorithmic dialog system is and for his all the potential that's bethesda's awful radiant quest system has, no one likes any of those games for those features. If Arcanum was entirely constructed with procedurally-generated a dialogue with a procedurally-generated story and so on, it simply would not be the game that everyone loves. For the same reason you wouldn't care about reading a book entirely written by an AI so too would you not care about playing a game entirely designed by AI, there's just not a intelligence there that can create a compelling narrative and even if someone found a narrative created by an LLMto be compelling, what would be the point because it's not made by a creature with intelligence! No narrative created algorithmically can ever mean anything because it wasn't created with intent it's just an engineered version of the average input of a bunch of different stories that you fed it from a bunch of different places, there's no purpose there's no subtext there's no anyting *it's just the narrative version of a pink goo back and be excreted whose only purpose is to save money on hiring writers.* Personally, I just don't know why anyone would care about engaging with any of this content! People find no man's sky relaxing because you're just drifting endlessly through the empty void, but why would I care about a town of gnomes besieged by evil dwarves if I know that it's impossible it will ever mean anything. Talking to someone else about this game would be as interesting as talking to someone about a dream you had but without even the interest of psychology to motivate it. The best case scenario of companies using this technology is that we get an endless amount of sludge output from whatever AAA games studios triggering a massive financial collapse when the AI bubble bursts as everyone realizes nobody cares about playing a game that was was made by what we now call an 'AI'. *What it all boils down to is that nobody wants to engage with media that no artist actually created.* So all your senarios with the drow in some random ruin raising a pitch or what have you are compelling things to base a game off, but there has to be more to it than *JUST* the baseline idea but with AI that's all you can get. We'd all be much better off with the same amout of effort going into people making actual games with actual stories.
This sums up my apprehension with the allure of "infinite" generated content pretty well. Human brains are evolved to expertly deduce patterns in everything, and that trait is ultimately what will hollow out the sense of novelty and expose the recurring tropes of the algorithms even if they can theoretically be much more sophisticated than anything we've seen in games with procgen thus far. Not having a satisfying narrative arc in generated quests/experiences is another good point, which makes us view this stuff more as "content" in the pejorative sense that the term is garnering nowadays. The manner in which procgen is integrated into otherwise hand made worlds is another point to consider. I think it's best kept rather small in scope like randomly generated loot in containers or character inventories which is already common, or otherwise have anything major be reasonably cordoned off like a particular dungeon that can be generated/infinite for some lore reason. Having procgen elements be integrated more deeply into the otherwise hand crafted world I feel creates too much of an awkward juxtaposition when it inevitably becomes easy for the player to tell which parts are designed by humans and which parts are entirely systemized by an algorithm.
Having mentioned Skyrim theres a set of faction mods that work together a bit like this. If you ignore a faction it will grow and expand. They will independently fight other factions if they come into contact. You need to decide if you want to tackle a faction early before they grow too powerful or let them take out others that you like less. You can be part of one faction or another. They use existing locations but i think they can expand into mod added locations. There are also mods that procedurally create floors in new dungeons (although they didn't ever work that well for me)
These forever games you are talking about in such detail sound amazing i always wanted a game that i could keep playing indefinitely add mod support to that kind of game and it would be a infinite game indeed , i hope you will someday work on a forever game as your magnum opus tim
oh yea, if you want to look at a road system like what you were talking about, i think Kenshi did something very very close to it. you can even see them in the Shift-F12 menu in-game and place your own, which will alter the ground texture and guide pathfinding for both NPC patrols and your own characters, which I have used to fix navigation issues in and around my own bases.
when i saw how good chatgpt has become, the first thing i thought of was the "tell me about" feature in the original fallout, but so powerful that each player would have unique coherent conversations with NPCs
Have been considering some of the techniques you talk about for filling in the world map but I run into seemingly irreconcilable problems when it comes to intention. It seems obvious that procedural generation can create experiences, but I don't know how you direct any of that content in a thematically coherent way. How would you address the idea that you want to maintain some kind of set of central through lines to your story and what lens it is going to provide on the world with the fact that the AI is just going to ... do stuff? Anyway, thought provoking as always. Much appreciated!
my gripe w pg content is that it feels like its wasting my time bc theres no intention behind it. i dont mind pg maps and terrain, im all for anything that reduces file size and frees up devs time. but in narrative driven rpgs, pg content, even if integrated seamlessly (and thats a tall order) feels pointless. i was much more forgiving of this when i was younger and had a lot of free time, but now that it takes effort to carve out time out of my day to play games i want to get to the meat of it, to feel like the gm is weaving a story that makes sense and has meaning, not shuffling cards and rolling dice w no intent. pg places will never have an overarching story w a lesson, w current ai anyway. all that said, pg works brilliantly in sandbox games like rimworld and minecraft. starsector tries to have the best of both worlds w a static core words and pg the rest of the map and to their credit it works seamlessly, but only bc the narrative and mechanics are made w pg in mind i think. a game like the og fallout would be very frustrating to play if i had to slog though pg towns and caves to get to the actual story, never knowing whats parts were made w intent and which were pg, also longer.
I think good procedural content can be said to have intent systemically. Like random encounter #139 doesn't have designed intent, but the designer DID have intent in making the world contain those types of dynamism. GTA would be a lot emptier without random pedestrians and traffic spawning all the time for example.
Fallout 4 does that to an extend. Obviously, not the way Tim described, but I played it for thousands of hours on the same savegame after handmade content was depleted filling the gaps with random encounters of 2 or sometimes 3 factions clashing with each other that you can partake in and in the end sell the stuff that you've looted from that encounter or leave for yourself. Also, many settlements and POI's would get repopulated after some cataclysmic event happens there after a couple of in-game days.
I like to think the implementation and design aspect for procedural generated content is very interesting. But as you say, it's still has a long way to go, especially for RPGs. For me it's still not much more than filler. In the main mission of Starfield i was send to two different procedural generated planets, to the same building which had the same layout and same corpses lying at the exact same place just to get an item. In addition by randomly exploring another procedurally generated planet i got into that same building again. I didn't enjoy that at all, i felt that was incredible lazy. Especially when this is part of the main mission. But for side content on the other hand, I was a bit disappointed that FromSoftware didn't include procedural generated dungeons into Elden Ring. They did made a decent first attempt in Bloodborne with the Chalice dungeons. I also like procedural map generation in strategy games like Endzone - A World apart, where you could even share the seed with others.
I think my general stance is that, I personally prefer the idea of handcrafted content not in fear that procedural content might be worse, but to allow that sort of building of knowledge of the game. Learning locations of items and npcs and interesting locations. Of course even my favorite game ever, Skyrim, does that a ton. But the idea of procedural content being the primary source of exploration makes it feel underwhelming. As tmgreat as the algorithm might be, id always have the thought of "this was randomly generated, it doesn't have much to offer besides resources and xp" In most games it's little issue but in elder scrolls for example, knowing that majority of locations you might encounter don't offer much if any new world building makes it a lot less inticing.
Same. I love my Lore and Narrative in games. I like that in some franchises they mention a far off place that you never get to visit or a major NPC that is talked about but never seen and so on. I like to look at maps and see towns and cities that we never get to explore but they have some in-game Lore behind them. I feel ProcGen towns and cities would take away from that for me.
Shadowrun for the Sega Genesis I think is the game that best illustrates the potential of this idea. It doesn't have procedural content but the "random" missions you get from the different Mr. Johnsons in the city are the perfect frame work for a procedural rpg in that style. Its honestly what I was hopping Cyberpunk 2077 would be, but they didn't go for that rout, instead putting a bunch of static encounters all over the place. A realistically sized city could have procedural quests the send you to streets and buildings that would take too long for a quest designer to come up with. The main problem is how do you make those quests stand out and not just be boring copy pastes just in a different location. Shadowrun trickled out different types of quests the higher class the Mr. Johnson you got them from. And honestly you could just repurpose the story quest missions into a more stripped down version for the procedural content, so you could still do escort/courier/hitman/kidnapping/bankheist/corporatesabatage/ect. missions without having hand crafted cutscenes and character dialogue. I could point to a bunch of other games that have elements that would have fit into Cyberpunk like the Kowloon section of Shenmue 2 having explorable and rendered apartment buildings that could be filled with procedural content, or the faction system in GTA letting you get reputation with different gangs/clients giving you different story outcomes with which one's you followed in your playthrough. We could sit here all day talking about what that game could have been (and i still really enjoyed it for what it was, just not what I envisioned it could have been), but I want to get to the real point. Microsoft has the console publishing rights for Shadowrun games, and Obsidian is now owned by Microsoft. Tim, you have to come out of semi-retirement, and get Microsoft to fund an Obsidian developed Shadowrun immersive-sim RPG with procedural quest and content generation. The possibilities are endless, with that frame work down you could expand into multiplayer having teams of users forming shadowrun squads to complete missions in co-op. Then you could have teams compete against each other in competitive asynchronous pvp, Then finally combine the whole thing into an mmo that lets you choose which cities in the world to live in and engage in those activities after all the groundwork has been laid over multiple games/expansions. The people need you Tim, please.
I haven't played it, but game called Pine has something like that with factions. Things change. And for existing stuff in games like Skyrim and such.... Make furniture movable or destructable. And Passwall.
I'm learning C, Godot (GDScript), and Game Maker (GML). I've been toying around with the idea of making a simple Catan clone. But now you have me thinking: make the Catan game be the procgen creation of a map where gameplay happens. I mean if you think about it, if someone takes the only spot available that connects stone and brick resources, and you needed those resources, you're going to be mad at that player. Instead of players being mad at each other, this can be used for NPC backstory, detailing why clans rival each other. Obviously would have to make a lot of changes so it's not literally Catan, but it can still be Catan at heart.
I agree, we only just see the very tip of iceberg of what advanced text generation models can do, within next 15 year I can imagine someone would figure out ow to add a "ever expanding dlc" to any story based game were player simply chooses the direction they would wish to go and computer tries its best to create custom designed content on the spot that will never repeat or appear in anybody else game play.
I'm not a game developer but "forever games" has been my ultimate pipe dream and is exactly how you describe it. The closest I have is Minecraft and roguelikes, but these are entirely procedural.
Can you talk about the music of Arcanum? That really brought the whole game together for me when I played it as a kid. I ended up using it as a sample for some music I made probably a decade later. It just ties everything together and builds so much atmosphere.
Not sure how you'd feel about it but I'd LOVE to see you collaborate with some Fallout modders to either make a new game or an overhaul implementing some of these ideas!
I love the idea of a system that truly keeps track of things that have an organic place in the same sandbox the player is in. My favorite example is a bandit camp that raids another one and absorbs all their wealth, and everything in the world kinda just does it's thing like that. Then there's the more literal ecosystem, bears will need to hunt to survive etc. and generally do cool bear things in order to be a bear, not a lifeless npc running off a simple script.
I prefer a shorter game with a handcrafted content rather than a game with a 1000s of bland and samely quests. Even some handmade games can be too long for their own good, there is no need to add some subquality content to the mix. The only way procgen can improve the gaming experience is when it enhances the simulationist aspects, like mentioned in the video factions wars.
Hey Tim, are you sure 'Forever Games' are a good thing for the industry? I think there's a lot to consider in terms of how companies will use it to keep players hooked, and only ever playing their game, or pouring millions of extra dollars into trying to keep a game chugging along when its time in the spotlight is over. There's more I'd say on this, but I'd be opening a massive can of worms.
procedural generated AI items, descriptions, even 3D model if you have a 3D game. Infinite unique looking items/armor/weapons. What we need is proceduraly AI generated quests based how you interact with npc's, if you can add voices too and event to that even better. In the future games will be 90% only AI made. There was a demo from nvidia, showing what AI can do, and was a guy talking ingame through the mic to the npc, and the npc was replying and based on game settings and what you were talking with the npc he could give you clues even missions/quests that was generated just for you
I believe that Handmade and Procedural Content could work well in a cohesive fashion with each other. This is not to portray a full scope, but potentially a method of how Handmade and Procedural Content can coexist with each other. The Handmade content provides the initial structure of the content within an umbrella of rule sets. While the Procedural Content is generated inside of that structure. That way the design can still have some oversight of the generation to some degree and can iterate or balance it because it lives within a defined structure. Perhaps an analogy to convey what I am trying to illustrate is this. For you mmorpg gamers, you know the concept of themepark and sandbox? Think of this more as a sandpark. Something like Everquest. The handmade provides the park portion and the procedural provides the sand portion.
I find the idea of "forever games" and how they're being thought of rather strange as I was under the impression games you can play "forever" already exist. In fact, the specific games mentioned, Fallout and Skyrim, already hold that distinction, partly due to the modding community's love for them. They've become classics in that sense, but just as the book, Lord of Light ,doesn't change in the least when Tim reads it every year (Tim does), neither do those games. The games themselves remain the same. It is we who keep coming back to them again and again..
The thing is that as a player as soon as I sniff that a quest is a procedural one (like the Radiant quests) I don't want to spend my time on it. With all the handmade content out there why would I want to spend time playing something randomly arranged by a machine with an unlikely grasp of quality?
The "action game with RPG elements" description didn't come about by accident. "RPG elements" are just that: half-assed and generally unteresting "elements" that lend nothing to gameplay and come across as afterthoughts at best. That's the way supposed "RPGs" in the industry at large have gone and I've not seen any hint of a course correction coming about anytime soon.
Check out Shadows of Doubt, it kinda matches what you described in the video with the procedurally generated NPCs, routines and conflicts/crimes/quests.
I know I'm a little late with this one but I was listening and thinking about it. I feel like the problem with current procedural generation in games is that it makes it all feel disjointed and gamey. Hard to convey the feeling I get from current games trying this. Or at least games with procedural elements that I have tried. My god Tim every idea you had here would neatly solve that disjointed feeling. Would really love to see this
That's a cool ass idea, using the Pokemon style rival system but it's a proc gen player created "antagonist" who builds their own party as the "protagonist" party advances. Twist, you can swap sides at any time and the AI takes over the opposing side using the choices committed (either ai or player made ones) as the seed for continued proc gen. You can play the whole game as 1 party against and AI, or you can take over as frequently or infrequently as you want. Build your own big bad.
If you look at the updatws for Unreal Engine 5 showcase the procdeual generation looks nearly identical to handcrafted content. Procedual generation has the potential to match handcrafting and if it can do that with a massive scale that will change gaming forever.
never ending / xp grind / fetch quests could be really unique and very different each time. Maybe an interesting big story unfolds, while you visit a place you have to steal something from and then you are in the middle of a big fight. There is a llot of potential for such games.
Is there a way to access those GURPS generators today? I'd love to fiddle with them, possibly try to recreate them for coding practice. I watched your videos on those GURPS generators when they came out but I don't remember if there was a way to find them online
Procedural generation seems to be trying to solve another kind of problem: cutting down on the time needed for coding in general rather than making new content. Speeding coding up seems to be the true problem here, and procedural generation at this time is a blunt tool to this end. The procedural problem also highlights the tug of war between quality and quantity in games. Either way, the clearer an idea the developer has for the way a game should be, the less there is the need to make quantity rather than quality.
Hi Tim As you mentioned Barks in this vid I've always wanted to ask... How come the same bark gets said for different characters using the exact same words even by different actors? Is it a - we have to pay the writer more if we ask them to make variations - type thing? Can't you ask the voice over artists to improvise some alternatives in the booth etc? (or does that come with costs too?) Anyhoo - thanks for all your wonderful stories :)
Universe, with heavy proc gen, sounds like Elite Dangerous. IIRC David Braben is hugely into proc gen. In honesty, I haven't played that game myself, but I was devoted to Elite as a child. You've totally sold me on proc gen for the likes of Fallout. I love the idea of the modding community getting their hands on the proc gen engine and coming up with ever more different ways for it to introduce emergent events. Funny you should mention AI. I was wondering today what your thoughts are about that, so it's good you mentioned you've done a video on that. That's next on the watch list!
the Nemesis System is another kind of Procgen done really well.
I loved it. Need to go back and play some shadow of war now. Nice.
I'll never get over the fact that WB patented it and has done absolutely nothing with it since. A horrible decision that is going to completely halt any innovation in one of the best new procgen systems for decades.
i always see you on lyle shnub videos, you have good taste
its implementation leaves much to be desired in both games it’s featured in, and we’re never getting another one so it’s essentially a dead mechanic now in terms of innovation potential.
Of course, on the indie side, mcdickie games were using it before and are still using it now. And even though I really love wrestling empire, those games are more akin to a really good shitpost in terms of production value.
@@ElvenSonic "essentially a dead mechanic now" IIRC there's word that WB's in-development Wonder Woman game builds upon Shadow of War's nemesis system.
But also what I've heard from people who look at the legal side of industry stuff is that the patent is somewhat vague. Maybe, for all we know, that could mean a deeper purpose of the patent is more about deterring other studios via the risk of *timesink* from legal proceedings, rather than WB assuming they've got actual assured court protection from anyone else being allowed to build upon those mechanics. Could help explain why one of the only "big" games to include something approximating the nemesis system was Assassin's Creed Odyssey, since Ubisoft is a powerful enough juggernaut that they probably already have in-house lawyers ready to go.
There is merit to something being finite. It can be completed, it can be exploited/predicted, you can discuss it with other people. That feeling you feel during New Vegas, the melancholic feeling of this is it... it's a powerful and addictive feeling. So many games/movies/series nowadays want to go on forever... but finishing of something and finishing it well is something the audience secretly craves and the investor laments.
A good ending can really seal the deal on something that was already awesome. And so few story writers get it right or even get the chance to write it.
Case in point - True Demon Ending for SMT: Nocturne. What an amazing game ending, you ascending and becoming a champion of the mightiest being in the game setting.
This was then made canon and your main character reappears throughout the series as final extra boss fight to clear, if you're masochistic enough to try.
I wonder how much experience Tim has with games that rely heavily on procedurally generated content, like roguelites and Dwarf Fortress and such. In my experience procedurally generated locations and characters get stale really fast, once you've seen half a dozen permutations it stops being interesting. On the other hand procedural generation that affects the initial faction dynamics does make playthroughs feel more fresh, like say in Mount & Blade some faction starts snowballing or in Dwarf Fortress your biome and neighbours affect your game a lot. So the smaller scale stuff is a lot better when hand crafted like characters, items, exploreable locations, but the larger scale things like map level biomes and faction power dynamics do benefit a lot from randomization.
I think part of this is just that the majority of designers doesn't really think about the generation too much. Bashing noise functions together to generate landscapes and dungeons is a very surface-level implementation of procedural generation at the end of the day. A true high-concept randomized game would have building blocks designed from the ground up to lean into generation, so that you can get those highlight rooms and encounters and everything doesn't feel the same. Of course, letting your generation have peaks like that runs the increasing chance of generating imbalanced or unsolvable runs, so there is a lot of consideration going into this beyond turning the RNG knobs up. But imo, most current procedural games play it very safe and use the technology just to provide comfortable stream of content instead of being interested in pushing the possibilities.
Dwarf Fortress sounds like a forever game
Ironically, Dwarf Fortress generates the smaller details, but it's larger structures, like civilization types and different groups of semi-random monsters, are almost like premade content. Civilizations are a lot more elaborate than procedural content usually is, but, for instance, goblin civilization has the same structure, same with every dwarf or elf civilization, and then they're rearranged in different positions in each world. Meanwhile individual dwarves or places on the map have a lot more flexibility in their generation.
Exactly what I was feeling. I have a lot of playtime in Minecraft which is built entirely on procedural content, and it has sustained an active player base for a really long time. But even with this "infinite" content, the developers still (rightly in my opinion) feel the need to keep adding more features and content. It seems like procedural content can take just as much effort to generate meaningful playtime as hand crafted, at which point it comes down to player preference. I usually start to notice the patterns in procedural content really fast and it lessens the experience for me, while others are perfectly fine to play them for thousands of hours.
The real difficult thing in procedural content is making it varied enough to remain interesting for a long time. Humans are great at noticing patterns so we're really hard to keep immersed with content generated through a rigid set of rules as those rules start shining through brighter and brighter with each time you play through a piece generated with them.
I think my main issue with a lot of the games I have played that rely on procedural generation is the eventual feeling that I get that it doesn't actually matter where I go or what I do.
If the proceedural generation is as likely to generate a quest/mission if I go North as it as if I go South, it doesn't matter which way I go. If I know that any given cave or dungeon was generated with a roughly fair difficulty based on my level, and will give me a randomly selected treasure appropriate for me, then it doesn't matter which one I go to.
If there is no real central quest or mission, then exploring the world and gaining power really has to be fun in its own right, and while sometimes I can get into a groove with a game like that, more often I find myself asking why I bother.
That's a good way to put it, and I think is the key issue with Starfield. Without systems in place to force me to live with my decisions, nothing matters.
A lot of proc gen is just sterile, the distinction between going north and south doesn't have to be so meaningless. I think this is an area of games we're yet to see the best of, more so than most other aspects.
@@Gnurklesquimp2 Agreed, I think there is a lot left to be attempted involving more robust procedural generation but it requires an attention to detail, worldbuilding and a development plan that incorporates procgen as a cornerstone from the beginning that many developers are going to be wary of for fear of not releasing a successful game.
There are very few games out there with handcrafted content that take things like the emergent impacts of physical geography and socioeconomic realities into serious consideration when designing their worlds (one of my big problems with bethesda games even though I love them, they feel like disney theme parks more than living worlds) let alone developers designing robust procgen systems around such things.
@@sieda666 That last thing you covered is something I often consider. AI doesn't have to be lifeless, in fact, it can bring life to complex things we struggle to by hand. This would already help with mostly static worlds, but I think we're gonna see worlds that live and breath through ecosystems in every major way, something I also wished for with every TES game I played.
Hell, some of the budget paying for people who'd otherwise be making things by hand can go to experts that merely review the AI's work. Not as simple if it generates stuff on the spot in playthroughs, but otherwise doable if you give them some tools to make it faster.
My biggest gripe is particularly focused on the latter half of your comment. If a game has really good gameplay, I couldn't care less if the dungeons I'm running are randomly generated and predictable, because it's the gameplay that's keeping me hooked. But that's rarely the case with games that rely on procedural generation. I loved snap map in doom 2016 because it meant I could play the game more, and in new ways. But it wasn't procedural. If it had randomly generated levels that went on forever, I wouldn't care. The gameplay is just that good.
Fallout 4, on the other hand? Its gameplay isn't bad, but it isn't good enough to justify playing for the sake of playing, in my opinion, and I would avoid the radiant quests whenever possible.
My point being games that have procedural generation rarely have the gameplay to back it up, and gameplay that is just that good that you'd wanna play forever always seems to be in games that are seemingly over wayyy too quickly.
I think binding of Isaac and ziggurat are kinda good examples of procedural content done right. But those aren't in the same wheelhouse as something like fallout or skyrim, the random generation is literally part of the gameplay, it's not there to extend the lifespan of the gameplay.
As others have mentioned I feel like the type of procgen being described by Tim is being pioneered by rogue likes. Dwarf Fortress, Cataclysm: DDA, and Caves of Qud come to mind
Definitely, was thinking his description of a fully-procedural Fallout sounds a lot like CDDA.
I like the Dwarf Fortress world generator. I prefer generators where the final product is a result of a sequence of rules based inpainting, with the setting sliders making use of the noise, instead of a simple fitting together of prefab tiles, and DF using volcanism and rainfall erosion to generate mountainranges, which in turn will inform civilizations is very appealing.
Wayward Realms made by the guys behind Daggerfall and a few other people is going to have procedural generation that accomplishes crazy things like what you’re talking about. They’re planning on making it into this interconnected world with all kinds of different things constantly changing how everything interacts. It’s pretty crazy what they’re planning. There will also be a bunch of handcrafted stuff to ensure the game still has a soul.
You think it’s ever coming out?
I certainly hope so. But, I will admit it seems doubtful. Plus, what they’re doing is VERY hard to pull off.
In S.T.A.L.K.E.R Clear Sky (one of the least liked Stalker games, which I think is unfair), it had this _really really cool_ faction wars system where NPC squads would wander around and interact with one another. Bases would get attacked and taken over or maybe the attack was repelled, but every interaction was signalled on your PDA so you could head over there and help out one side or the other, or you could just wait until it was all over, pick off the survivors and loot everything
It was amazing but so few games use that system
There was a mod for Kenshi which did something similar and I crave that in open world RPGs
my brother and I also use the term forever game to describe theoretical games with worlds that are as close to living as we can imagine :P
If you like Clear Sky which I liked too, probably the best proc-gen, open play STALKER mod right now is GAMMA, look it up on youtube, it’s based on Anomaly but with a lot of annoying ideas stripped out and core gameplay loop enhanced. You can work with any of the factions from the Free Stalkers, through Duty, Freedom, Mercs, Ecologists, etc, with rewards that tie into the factions. In a lot of ways it plays like Mount & Blade: STALKER, and I love it!
Edit: Also unique gameplay modes like Azazel Run, where if you die you become a random stalker in the zone of any faction with whatever gear they have, faction relations. It also still has the base Anomaly quest line if you want to do a main story too.
It happened too much tbh
GTA san andreas had turf wars between gangs that happened periodically /dynamically, and you could choose to participate, added quite a bit of life to the city.
i guess this is just a trope of the genre to an extent but i'm a bit put off by the idea that the whole point of this is to just make a game as big as possible. as i near my 30s i'm becoming increasingly aware of the reality that my time is most certainly NOT finite and i don't like the idea at all of just filling all that time with just one piece of art. there are so many incredible video games (and movies and shows and music and books) that i want to get to experience in my lifetime and it makes me pretty anxious to think that many of them are RPGs that will take several hundred hours each. i understand your enthusiasm tim and i'll admit there are no doubt a solid handful of games that i love enough to be able to play perpetually if i was forced to, but i also have the choice of playing many many more wonderful games instead and broadening my artistic horizons and i'd much rather do that
I came to really enjoy Daggerfall for the scale, it really felt like my own story because of how lost I got among the many cities and towns I based out of. I tend to replay handmade games more, but I am looking forward to the Wayword Realm's upcoming title that explores these large scale worlds.
In terms of quest, story, and even NPCs being generated by Ai; I'd have to say that it should be used to assist the existing experience. The main story, main hubs, and characters should be hand made. Games are becoming like books in many ways and should reflect intent beyond 'making money'. So they should have a crafted experience.
Wow, didn't expect people to be talking about Wayward Realms here. It feels like nobody knows about them and their game (I only found it, I think, by googling Daggerfall's original designers and what they were up to). I also liked Daggerfall, trying it for the first time a couple years back, but what absolutely destroyed my desire to play is the god-awful dungeons. They are a perfect example of procedural content gone wrong, in my opinion. Just soulless mazes of the same corridors and rooms, randomly rearranged over and over...
@@BuzzKirill3D I can't disagree with the procedural generation of the dungeons being not great. But when you consider the time for it it was fantastic. Thank god for mark/recall though. Daggerfall was almost 30 years ago and to me is still a cornerstone of procedural generation for areas. Honestly with Oblivion and Skyrim (and some of Fallout 3 and 4) I g to a little tired of the "dungeons" almost always having a shortcut back out to streamline the handmade dungeon experience. The sense of exploration that those procedural dungeons had was great, you could get just utterly lost in them for hours.
You were saying "imagine" when talking about procedural areas and quests, but there's already a game that does that. Dwarf Fortress is a great example of both the strengths and weaknesses of proc-gen: it's infinitely replayable for the same reasons you mentioned and more, but everything in it lacks the detail of a more handcrafted game. It's also been in development for an exceptionally long time, although a lot of that was caused more by feature creep than issues with the engine.
The most interesting recent exemple I've seen of truly good procedural generation I've seen is "Shadow of Doubt". It's a detective game with procedurally generated crimes, in a procedurally generated city district, with hundreds of generated citizens. And it adds a very important ingredient to the procedural generation : *simulation*.
It simulates what everyone is doing, they go to their jobs, leave notes etc... It is very finite in what it represents: the city is small. But it is very dense, and it feels very coherent. Because the rules that place items in the world make a lot of sense.
I like No Man's Sky, but everything in the world is very temporary, which makes it harder to care about it. The different elements lack connections, two individuals will never communicate in any meaningful way.
Shadow of Doubt is still in EA, right?
Personally I hope someone makes remake of first two Fallouts filling it to the brim with procgen events/things which would however make complete sense....like Tim said.
Although my realistic self thinks that at least for first few years of procgen boom the games working on that mechanics/algorithms are going to be bad and boring...with few exceptions though......since the procgen stuff will be only as good as the core mechanics/art/story created handmade. As long as we will have AAA studio releasing bland and shallow games then not even procgen can change it into masterpiece.
All your examples sound decent when just initially brainstorming, but making this all work while being fluid, making sense, and not being obviously prefabricated cookie-cutter looking assets is really tough. The explosion of procedurally generated games in the past decade have been interesting, but all of them that I have played, I instantly wish they just hand-crafted the entire thing. The single game that I have played where it works is Binding of Issac, but even then, that is a relatively basic system that just happens to work based on the strength of the core gameplay loop, as opposed to the complexity of, say, a randomly generated Metroidvania or open world RPG needing to nail exploration and proper item progression.
You should try Rimworld. I consider it the most resounding success of PG content.
Minecraft.
Also, the idea with procedurally generated quests is just one quest you can do 50 times. These are not 50 new quests, not 50 new stories, thats 1 quest you do 50 times in different areas. Its gets really old, really fast in my opinion.
Especially when very few variables change, like in the Bethesda style radiant quests, where two names and a location pointer is the only thing different between iterations.
Hey.. glad you’re here.. we lost our friend in the cave 400m away and need you to go pick him up. Hey wow.. so glad you’re here.. my group got stranded can you walk me 1500m to my starship?
This is like half of the random encounters so far in 60 hours of SF
The point is it could be so much *more* as it's built up and built out. Theoretically there's no reason it couldn't generate entirely new types of missions, generate actual stories with plots that take you to specific places for a reason rather than just randomly putting a McGuffin somewhere. It wouldn't be easy, but I think there's an entire ocean of untapped potential because devs stop at "welp, it technically works." and anything beyond that requires a lot of work, R&D and money.
@@Xbob42 " generate actual stories with plots " -- I'm with him --> "'Each time, the team collectively believed - believed down at their CORE - that this was possible,' [David] Gaider said.... 'Just within reach. And each time we discovered that, even when the procedural lines were written by human hands, the end result once they were assembled was… lackluster. Soulless.' [David] Gaider believes developers - indie and AAA - and executives will still continue to push for AI in games, but that it will only result in the same conclusion BioWare drew and end up being a colossal waste of time and resources." And perhaps even put off enough RPG players to make them finally sit up and take notice. Of course, that won't happen until and unless their sales drop off even more sharply than they already have.
@@lrinfi A team failed, ergo impossible. Got it.
The really interesting aspect with procedural content, is that if it's sufficiently complex and dynamic it can even surprise its own creator.
In a lot of ways I usually look at it the way I look at Braid etc., they think of a mechanic and see what can be done with it through a process of imaginative exploration, the best mechanics they find get turned into incredibly elegant games.
I watched a talk about dwarf fortress
I think was a gdc talk
The cats were dying and they creator had no idea why
but apperently they were walking in bars and then cleaning their paws and then dying from alcohol poisoning.
th-cam.com/video/6yWf6BHqiWM/w-d-xo.html
I appreciate your enthusiasm on the topic, but for me procedurally generated content is just inherently uninteresting.
I would just much rather experience a story that some real person has put thought and emotion into.
The only example where I can appreciate proc generation are rogue likes. But even there I have grown tired of the absolute sea of roguelikes out there..
Unrelated question, but what is the story with the Captain's hat in your profile picture? Any major significance or is it just a nice hat?
It was a hat I found in Charlie Staples office a few years ago. I wore it for a few minutes and thought it looked fun, so I snapped a photo.
@@CainOnGames Ah yes, never underestimate the power of a great hat. Thanks for the response.
Procedural is the death of an adventure
There's plenty of adventures built on the premise of procedural uniqueness being at the core of the gameplay. Just because AAA hasn't done an amazing job of it lately doesn't mean it's inherently bad
6:50 While mentioning fallout, I always wanted a 3D fallout where you could travel long distances like daggerfall.
Always felt a 3x3 km2 Map was a bit small.
You keep smaller important areas hand-crafted and the areas between major locations procedurally generated.
Your idea is basically what I always wanted in a game.
sounds like a good plan for a fo1/fo2 remake, and obviously bring back the highwayman lol
@pira707 Definitely, for the long distances, it would be so good.
Proc-gen is one of my favourite features in games. Great to hear more of your thoughts on the topic! I love roguelikes and I'm working on my own using wave function collapse.
Thank you for your videos! This one in particular was a great discussion. Totally agree on the value add of procedural generation and the potential to add richness and narrative.
The problem with Starfield, from my understanding is although radiant procgen caves and buildings very much do appear on planets, they often repeat a bit too often because Bethesda hasn't got enough handmade content to make it feel unique. Examples are caves with the exact same skeletons in, or outposts with the exact same layout and NPCs in the same place. I also personally think the 1000 planets is a bit much at this time and they perhaps should have reduced the scope a bit to allow for more unique procgen radiant encounters.
It's probably impossible to make enough handmade content to make things not repeat.
Think about it. If in 5 hours you encountered the same location 4 times, then even if they doubled their handmade content you'd still come across it 2 times in 5 hours. And even if they quadrupled their handmade content you'd still see a repeat every 5 hours.
The truth is, you can't make enough handmade content to solve this problem on this scale. What they need is AI generated content but AI isn't going to produce Bethesda level content to rival their handmade stuff. And keep in mind that the more "radiant" a game becomes the more bugs can be introduced out from under the supervision of the team.
I don't like the idea of “forever games”.
I want the game to eventually come to some conclusion.
Sandbox is cool, but I really doubt it's possible to scale quality of random generation to the level of content made by a creative person.
That said, I don't mind a bit of it if it's a well organized system that takes a large set of parameters and turns them into small quests. It's always nice to have some side activity that isn't obviously staged (looking at you, cp77 🤨) and doesn't repeat several times in a row.
I mean you can have both, there is no reason a game can't have a campaign with a definite, satisfying conclusion and also infinite side content. A lot of the games that people really play for years on end though have a very small amount of content, but people don't get bored because the mechanics on their own are fun to enough engage with and improve at endlessly (e.g. fighting games, chess, tetris.)
"I want the game to eventually come to some conclusion."
The moment you stop playing is by all means a conclusion in itself.
@@fus132 I'd rather experience a story with a beginning middle and end then brainfuck myself into thinking uninstalling is an ending
Tim did you ever play dwarf fortress? (Or something similar like rimworld/prison architect). I'd be curious what your thoughts on them. Since i knew fallout and dwarf fortress i always wanted to make a game that combines them both... but then kenshi and rimworld came out so they kinda did that... (Kenshi is awesome too, btw)
You'd love Elite Dangerous :) It's a forever game with the entire galaxy to explore. It could use a bit more "handmade" or better procedurally generated story content, but it has a lot of the exploration features you were mentioning :)
I think some of the best procedural/infinitly-growing feature in a game was the nemesis system from Middle Earth, especially when you brought up new stronger people showing up in the game world it immediately made me remember how often the difficulty changed in Shadow Of Mordor based off of the way I fought my current enemies
Hi Tim! You're my hero, these videos mean a lot to me and make me smile every time I see them. They're also very damn useful for any game dev to check out. So, with that in mind, i'll drop in a question pertaining to my occasional struggles in development.
When stuck in a development rut, or any other applicable synonym of writer's block, what methods do you believe are most effective in remedying this issue? How do you avoid getting into a rut?
Thanks so much for all of your videos and help, as well as for taking the time to read our questions and such. You're amazing.
Some of these have been implemented, in more minor manners, in other games.
Like for example, when you describe wanting the world to feel alive, I remember black and white having an almost "ant colony" vibe to it, where you play as a god, you can direct your worshipers to certain points, and you get to see them essentially do it in their own way. I remember settlers 3 did something like this too.
The idea about a "special" NPC arising at some point to challenge you, that was actually closely implemented by the shadow of mordor/war games, the nemesis system.
Essentially when you'd die to a random low level enemy (even by accident), he would get promoted and in the game rise ranks, he'd become a commander, eventually if you kept dying to him, he'd become a warlord and keep increasing in power, both physical and in terms of the armies he could command.
thank you Tim, i was really reluctant about procedural content until i saw this video, you made some good points and i'm actually hopeful for a few games that'll come out soon, great video as always.
If I understand you correctly, you are talking about a living world which evolves within your setting for a player to immerse themselves in.
I love that idea. For me, that’s the holy grail for an mmo rpg. An mmo rpg where the world does not revolve around the players, but the players nudge the world in a direction.
Ngl Tim, you are so inspirational.
I would love a podcast !
Would love to hear your view on the current Unity debacle, and maybe any similar history you have come across.
Tim, I love your take on procedural content! We've barely scratched the surface of its potential for fascinating gameplay experiences. I hope to help remedy that as I use it extensively in the action-RPG I'm (slowly but surely) developing. 😄 Anyway, thanks again for sharing your thoughts!
Tim, I just want to say that thank you for using this platform to talk about game design. I have been formally trained in graphic design and the fine arts but recently I have been making my own game. Your philosophy or even just what is made into a game has really opened my eye to what even it takes to make a game. I just want to make sure you know that you really have made an impact in my design journey. Thank you. Also are you working on anything currently?
I like the combo platter, hand crafted and procedural together.
I love this topic. I could watch an hour of Tim Cain just listing out ideas.
I think the epitome of these being used together in harmony, is Ultima Online. That's one hell of a game; static cities, dynamic in-world player housing. Usual random spawns, able to tame any entity in the game and it becomes unique.
Arena Dagger fall comes to mind. Imagine a game with core content, but also has stuff like a dnd hex crawl, with random generation , random interactions. merge the rogue-like with the RPG on a complex and never repeating scale.
Diablo 2 had procedurally generated maps that work well and that game came out a long time ago. Procedural content can and has worked. Developers have also experimented with it unsuccessfully in the past but I think all that effort is worth it to realise its potential.
It really depends on how it well it's implemented, some of the best examples I have tried are Minecraft, Sid Meier's Civilization, and Mount and Blade. Some of the worst would be Bethesda's radiant quests, but maybe the simplicity of the gameplay is to be blamed there. If stealth in Skyrim had the depth of Metal Gear Solid, those Thieves Guild radiant quests might be fun.
The only game I have seen that has procedural generation that is rich enough to create interesting stories is Dwarf Fortress. I could tell many stories of silly things that happened during my games or tales of how my fortresses ended or a story about the one time I tried to pump water and accidentally made a water slide of death.
So, this makes me think of Dwarf Fortress, which I see others have also mentioned in the comments. To me, procedural content can lead to sludge or to diamonds. In DF, it leads to diamonds. In lots of other games it leads to sludge. Why? Well, DF is set up as a history generator. The game generates a world with a mythology and a world history with detailed events transpiring between settlements and individuals. This means that what the game presents you with is not just procedural sludge, but people and monsters and places that have an evolved history and connection to the world that they and you are in. Things have meaning and a history. It's not just perl-noise in the shape of game assets or whatever. And ultimately, to me, a really big thing that I play games for is history. Sometimes, it's the histories that the game's creators have decided to tell; and sometimes, like in DF it's the histories that evolve.
Great insight into the potential of procedural content. I really enjoy well-done procedural content and believe it’s the future of RPG’s with sim elements like Daggerfall.
I recommend everyone here try the Fallout Fixt mod, so you can see a restored version of the original super mutant invasions. It has a similar dynamic feel to what is discussed in the video.
Thank you for the great insights Tim! I love the application of procedural generation when done mindfully and when being appropriately explorative so as to avoid excessive repetition! I think Arcanum's dialogue was successful in that.
Can you make a video on the procedural perk generation you mentioned? That would be interesting 😊
Have a great day!!
I cannot express how surreal it was watching this video and you explaining the same things I've been so interested in over the last couple of years. I've always wanted to create a "forever-game" version of Fallout, and procedural generation is so interesting to me that I'm currently shaping my senior project for undergrad around making a general-purpose procedural tilemap generator for Unity that will (hopefully) be able to include structures and dungeons and NPCs. I'm bookmarking this specific video for future reference if I ever have trouble explaining what I want to create to other people, and will be referencing it as a source of inspiration (even though it came out after I started my project). Thanks for the great insight as always Tim.
Bloodbourne has a nice procedurally generated area called the chalice dungeons. They go on forever.
I literally had similar thoughts about procedural generation, about what is possible. I plan to create very sophisticated algorithms for my space exploration and combat game, factions with alliances and feuds. I also thought about how to avoid making the generation too stale or boring. Because that is the core problem. I too believe that doing the algorithms properly is the solution to the problem. Doing bad algorithms is like doing a bad job in "handcrafted" elements.
I think it's just more challenging to make algorithms right, as they are quite heavy on math and logic.
One example of how not to do it: Assume planets can have attribute A, B, C, each of which make them interesting/relevant. Without them, they are useless to the player. Giving a chance of 10% for each of those to occur means the chances for 2 of 3 to occur is 1%, and 3 of 3 to occur is 0.1%. But we don't want the player to comb through thousands of planets to find one that is interesting.
Solution: Either increase the rate of multiple of A, B, C to occur higher, or create mechanic that allows the player to quickly eliminate irrelevant planets or to find relevant ones.
Honestly, procedural generation in SF detracts more than it adds. The main points of Bethesda games were exploration of handcrafted universes and complete immersion. First is torpedoed by procedural generation, second by constant fast travel and cutscenes.
Yeah, that's why I fell in love with Morrowind, back in the day, despite its flaws.
Sadly, the sequels failed to fulfill its potential...
I would argue that's Tim's entire point: Bad procedural generation is the issue, not procedural generation overall.
Also the handcrafted engine Bethesda used was so fucked up they couldn't do seamless travel when like 12 nerds in the UK did it years ago for their indie space game. The fuck is Bethesda even doing?
@@Xbob42Because rebuilding their engine from scratch (which is used for more than ONLY SF) just for seamless flight into planets is not worth the huge investment for a game that isn’t a space sim. NMS’s bespoke engine was specifically created for a single space sim game.
@@tronam Yeah I know, they made Starfield shittier because they didn't want to put in work or use an engine better suited to the game they made. You can say it's not a space sim all you want but good luck finding anyone fuckin' dumb enough to think the game is better filled with nonstop load screens and fast travel.
Classic Bethesda.
Well said
You are right. People in general do not understand the potential of proc gen, but rather focus on the limitations of past games that tried sell it in their marketing...
None of the procedural stuff works well when it comes to narative design and dialog. That's why in Minecraft, there is no dialog or quests. As soon as you introduce the narative aspect it all becomes boring, repetative, fake and stitches are revealed. So no, it's not a good idea to lean heavely on the procedural stuff in future Fallout games. Starfield is a disapointment in large because there are very few quests.
Procedural generation is something I've wanted to explore further, almost enough that I may have pursued a PhD around it. (I wanted to do a multi-node parallel processing cluster to run a Dwarf Fortress and greater level of world sim that constantly runs.)
The problem you touched on with current examples of procedural generation... They simply don't go far enough. A lot of the procedural generation systems I've seen out there are limited for one reason or another, and they simple don't go past the bar they need to. And a lot of examples front-load on initial generation... but do nothing later, which makes it easy enough for players to even subconsciously detect the pattern.
I think current gen hardware finally has plenty of processing power and storage capacity one average to really do a procedural generation properly. I can imagine layers of small scope generation systems being layered upon each other and the output laminated to a "base" state. Then, you take this laminated state as input another set of generation systems to layer on their own laminated base state. And you repeat for each category of meta-layer: Geology, Biology, Anthropology, Technology, Sociology, Political, and Psychology... So, you end up with a bunch of laminated based states that are like transparencies upon each other to project a new world to the player.
Now, the crucial thing is the ability to call upon the systems you used prior in the game to reprocess any deltas written upon the base states to generate new base laminate states and cascade the change up the stack. This feedback loop is what's missing in a lot of procedural generation systems out there. This is why stuff feels so stale, because it really doesn't change to what you do. If the player hunts a particular monster, it should start showing up less and less, until it's either extinct or moves to another part of the world where it doesn't feel threatened. If the player stirs up drama between two towns, I expect to see patrols around their borders and maybe even checkpoints on the roads. If I'm the personal grim reaper for a particular gang, they should either be sending hit squads after me or clear out when I show up.
For procedural generation to work, the world needs to react to the same degree it was generated. For anyone that's played a tabletop RPG, the best campaigns have always been when the world reacts and changes to what your characters do. It doesn't matter how it started, it matters how it changes. Procedural generation can allow for infinity permutations of change depending on the player, it's just not utilized properly.
To add, it has to change not only to people, but to environments as well. People say the problem with Starfield is everything is random, therefore nothing feels unique or immersive. The truth tho is that nothing feels cohesive or organic. When you get to a world with the pirate captain, you see a military base that was a secret lab to an opposing faction. You hear a monster got loose. Raiders have control of this world and are freely traveling amongst it. This tells a big story that the world needs to build around. Maybe the raiders have repurposed machines to the factions left over items to their needs. Maybe they made this beast its pet out there? This could be a theme that the procedural world builds around to make it feel alive. What I explained tho, this location is part of the story. Now you go to a random planet. It just has random stuff happening for no reason. Nothing feels immersive because the world isnt reacting to itself or the content it made for itself.
Imagine, that the world created a flower...a poisous flower...now its draining the nearb6y lands of water and creating toxic pools. The trees begin to be all deformed due to lack of nutrients. The people here now survive by living high up to escape the fumes. They harness the nature to create weapons since the outside world holds food that they need, that or use it for drugs to create a crime lord planet. This displays reasoning and logic with itself to create random worlds and experiences that react to the algorithm.
This is my biggest disappointment with these types of games. They half ass it. They forget they have to actually work to make it feel alive. Actually have to think bout how things interact.
I think rimworld and dwarf fortress really exemplify this sort of procedural narrative “forever story”
A lot of your points in the second half I would like to agree with because in theory they're great ideas but after Starfield I'm becoming more and more against it. I've already given up on Starfield purely because I'd keep running into procedurally generated environments, quests and NPCs and it was all just so bland and obvious. It also ends up unintentionally drowning out the hand-crafted elements of the game and pushing them aside. For almost every marked/pre-made environment in Starfield I kept wondering if it was procedurally generated or not and I mean that in a bad way (as in not a compliment to Starfield's procedural generation mechanics).
In my opinion writing and worldbuilding are the two core pillars of RPGs and having random generation take too much of the load off can ruin the whole experience. I have no interest in completing a revolving pool of generated quest types no matter how good the gameplay is. I don't think there's a way for a game to have that ideal amount of unlimited content and still be good. I'd much prefer games end and leave me wanting more than take shortcuts to try to give me everything and ruin the experience
The procedural generation does not always need to run on the players computer. It can also be used as a tool to pop out a lot more content during development. Mixing algorithmic generation, using handmade elements and doing a manual pass over the result / manually select the best generations.
12:38 I remember being absolutely blown away when I was playing Anarchy Online for a bit with how there was a literal floating city that would appear at random and my non paying newb self didn't understand how to get up there. It was awe inspiring. And this may be a bit of a stretch to say this is the reverse of what you were discussing but that game had a novel series attached to it and somehow players' actions would sometimes be a factor for what would get written and focused on
*The problem is that infinite in this context is still just finite with a little mask on it.* Any content that you design an algorithm to make is in effect really still just you making content. No matter what you do you can never design an algorithm to becreative because there's no intelligence behind it beyond what you are putting into it. And even if it was done the best I possibly could and you somehow created a system that could endlessly generate conflicts for some kind of reason for an endless numbers of places, let's say by running 100000 games of dwarf fortress in the background to create a somewhat interesting and seemingly unique history for anywhere you go, you're still going to be confined by the parameters that make up the algorithm. Let's say you even set up a large language model for each character in each one of these places so that they're all populated by unique NPCs with a history that can talk with unique dialogue in a infinite number of voices, so you have a infinite number of infinite places all with unique histories and characters that interact with one another all according to an algorithm. What's going to happen is that even if then somehow ends up being compelling at the start, it's all going to get incredibly see me with the same kind of characters having the same kind of conflicts that all talk in the same kind of way no matter where you go. So instead of just having one crafted game with a finite amount of content, now you have a averaged out smooth version of that game but stretched out over a hypothetical infinity. *That sucks.*
For as interesting as your algorithmic dialog system is and for his all the potential that's bethesda's awful radiant quest system has, no one likes any of those games for those features. If Arcanum was entirely constructed with procedurally-generated a dialogue with a procedurally-generated story and so on, it simply would not be the game that everyone loves. For the same reason you wouldn't care about reading a book entirely written by an AI so too would you not care about playing a game entirely designed by AI, there's just not a intelligence there that can create a compelling narrative and even if someone found a narrative created by an LLMto be compelling, what would be the point because it's not made by a creature with intelligence! No narrative created algorithmically can ever mean anything because it wasn't created with intent it's just an engineered version of the average input of a bunch of different stories that you fed it from a bunch of different places, there's no purpose there's no subtext there's no anyting *it's just the narrative version of a pink goo back and be excreted whose only purpose is to save money on hiring writers.*
Personally, I just don't know why anyone would care about engaging with any of this content! People find no man's sky relaxing because you're just drifting endlessly through the empty void, but why would I care about a town of gnomes besieged by evil dwarves if I know that it's impossible it will ever mean anything. Talking to someone else about this game would be as interesting as talking to someone about a dream you had but without even the interest of psychology to motivate it. The best case scenario of companies using this technology is that we get an endless amount of sludge output from whatever AAA games studios triggering a massive financial collapse when the AI bubble bursts as everyone realizes nobody cares about playing a game that was was made by what we now call an 'AI'.
*What it all boils down to is that nobody wants to engage with media that no artist actually created.* So all your senarios with the drow in some random ruin raising a pitch or what have you are compelling things to base a game off, but there has to be more to it than *JUST* the baseline idea but with AI that's all you can get. We'd all be much better off with the same amout of effort going into people making actual games with actual stories.
This sums up my apprehension with the allure of "infinite" generated content pretty well. Human brains are evolved to expertly deduce patterns in everything, and that trait is ultimately what will hollow out the sense of novelty and expose the recurring tropes of the algorithms even if they can theoretically be much more sophisticated than anything we've seen in games with procgen thus far. Not having a satisfying narrative arc in generated quests/experiences is another good point, which makes us view this stuff more as "content" in the pejorative sense that the term is garnering nowadays.
The manner in which procgen is integrated into otherwise hand made worlds is another point to consider. I think it's best kept rather small in scope like randomly generated loot in containers or character inventories which is already common, or otherwise have anything major be reasonably cordoned off like a particular dungeon that can be generated/infinite for some lore reason. Having procgen elements be integrated more deeply into the otherwise hand crafted world I feel creates too much of an awkward juxtaposition when it inevitably becomes easy for the player to tell which parts are designed by humans and which parts are entirely systemized by an algorithm.
Having mentioned Skyrim theres a set of faction mods that work together a bit like this. If you ignore a faction it will grow and expand. They will independently fight other factions if they come into contact. You need to decide if you want to tackle a faction early before they grow too powerful or let them take out others that you like less. You can be part of one faction or another. They use existing locations but i think they can expand into mod added locations. There are also mods that procedurally create floors in new dungeons (although they didn't ever work that well for me)
These forever games you are talking about in such detail sound amazing i always wanted a game that i could keep playing indefinitely add mod support to that kind of game and it would be a infinite game indeed , i hope you will someday work on a forever game as your magnum opus tim
oh yea, if you want to look at a road system like what you were talking about, i think Kenshi did something very very close to it. you can even see them in the Shift-F12 menu in-game and place your own, which will alter the ground texture and guide pathfinding for both NPC patrols and your own characters, which I have used to fix navigation issues in and around my own bases.
when i saw how good chatgpt has become, the first thing i thought of was the "tell me about" feature in the original fallout, but so powerful that each player would have unique coherent conversations with NPCs
Have been considering some of the techniques you talk about for filling in the world map but I run into seemingly irreconcilable problems when it comes to intention. It seems obvious that procedural generation can create experiences, but I don't know how you direct any of that content in a thematically coherent way. How would you address the idea that you want to maintain some kind of set of central through lines to your story and what lens it is going to provide on the world with the fact that the AI is just going to ... do stuff? Anyway, thought provoking as always. Much appreciated!
Those are such damn cool ideas. You should design video games
my gripe w pg content is that it feels like its wasting my time bc theres no intention behind it. i dont mind pg maps and terrain, im all for anything that reduces file size and frees up devs time. but in narrative driven rpgs, pg content, even if integrated seamlessly (and thats a tall order) feels pointless. i was much more forgiving of this when i was younger and had a lot of free time, but now that it takes effort to carve out time out of my day to play games i want to get to the meat of it, to feel like the gm is weaving a story that makes sense and has meaning, not shuffling cards and rolling dice w no intent. pg places will never have an overarching story w a lesson, w current ai anyway.
all that said, pg works brilliantly in sandbox games like rimworld and minecraft. starsector tries to have the best of both worlds w a static core words and pg the rest of the map and to their credit it works seamlessly, but only bc the narrative and mechanics are made w pg in mind i think. a game like the og fallout would be very frustrating to play if i had to slog though pg towns and caves to get to the actual story, never knowing whats parts were made w intent and which were pg, also longer.
I think good procedural content can be said to have intent systemically. Like random encounter #139 doesn't have designed intent, but the designer DID have intent in making the world contain those types of dynamism. GTA would be a lot emptier without random pedestrians and traffic spawning all the time for example.
Sid Meyer's Alpha Centauri.... play that over and over and over... I think I'll install it again tonight!
Fallout 4 does that to an extend. Obviously, not the way Tim described, but I played it for thousands of hours on the same savegame after handmade content was depleted filling the gaps with random encounters of 2 or sometimes 3 factions clashing with each other that you can partake in and in the end sell the stuff that you've looted from that encounter or leave for yourself. Also, many settlements and POI's would get repopulated after some cataclysmic event happens there after a couple of in-game days.
I like to think the implementation and design aspect for procedural generated content is very interesting. But as you say, it's still has a long way to go, especially for RPGs. For me it's still not much more than filler.
In the main mission of Starfield i was send to two different procedural generated planets, to the same building which had the same layout and same corpses lying at the exact same place just to get an item. In addition by randomly exploring another procedurally generated planet i got into that same building again. I didn't enjoy that at all, i felt that was incredible lazy. Especially when this is part of the main mission.
But for side content on the other hand, I was a bit disappointed that FromSoftware didn't include procedural generated dungeons into Elden Ring. They did made a decent first attempt in Bloodborne with the Chalice dungeons. I also like procedural map generation in strategy games like Endzone - A World apart, where you could even share the seed with others.
I think my general stance is that, I personally prefer the idea of handcrafted content not in fear that procedural content might be worse, but to allow that sort of building of knowledge of the game. Learning locations of items and npcs and interesting locations.
Of course even my favorite game ever, Skyrim, does that a ton. But the idea of procedural content being the primary source of exploration makes it feel underwhelming. As tmgreat as the algorithm might be, id always have the thought of "this was randomly generated, it doesn't have much to offer besides resources and xp"
In most games it's little issue but in elder scrolls for example, knowing that majority of locations you might encounter don't offer much if any new world building makes it a lot less inticing.
Same. I love my Lore and Narrative in games. I like that in some franchises they mention a far off place that you never get to visit or a major NPC that is talked about but never seen and so on. I like to look at maps and see towns and cities that we never get to explore but they have some in-game Lore behind them. I feel ProcGen towns and cities would take away from that for me.
Shadowrun for the Sega Genesis I think is the game that best illustrates the potential of this idea. It doesn't have procedural content but the "random" missions you get from the different Mr. Johnsons in the city are the perfect frame work for a procedural rpg in that style. Its honestly what I was hopping Cyberpunk 2077 would be, but they didn't go for that rout, instead putting a bunch of static encounters all over the place. A realistically sized city could have procedural quests the send you to streets and buildings that would take too long for a quest designer to come up with. The main problem is how do you make those quests stand out and not just be boring copy pastes just in a different location. Shadowrun trickled out different types of quests the higher class the Mr. Johnson you got them from. And honestly you could just repurpose the story quest missions into a more stripped down version for the procedural content, so you could still do escort/courier/hitman/kidnapping/bankheist/corporatesabatage/ect. missions without having hand crafted cutscenes and character dialogue. I could point to a bunch of other games that have elements that would have fit into Cyberpunk like the Kowloon section of Shenmue 2 having explorable and rendered apartment buildings that could be filled with procedural content, or the faction system in GTA letting you get reputation with different gangs/clients giving you different story outcomes with which one's you followed in your playthrough. We could sit here all day talking about what that game could have been (and i still really enjoyed it for what it was, just not what I envisioned it could have been), but I want to get to the real point.
Microsoft has the console publishing rights for Shadowrun games, and Obsidian is now owned by Microsoft. Tim, you have to come out of semi-retirement, and get Microsoft to fund an Obsidian developed Shadowrun immersive-sim RPG with procedural quest and content generation. The possibilities are endless, with that frame work down you could expand into multiplayer having teams of users forming shadowrun squads to complete missions in co-op. Then you could have teams compete against each other in competitive asynchronous pvp, Then finally combine the whole thing into an mmo that lets you choose which cities in the world to live in and engage in those activities after all the groundwork has been laid over multiple games/expansions. The people need you Tim, please.
Your idea of faction warfare reminds me how i like to boot up age of empires 2 or civilization and watch the bots fight each other, always amusing
I haven't played it, but game called Pine has something like that with factions. Things change.
And for existing stuff in games like Skyrim and such.... Make furniture movable or destructable. And Passwall.
Give this man inf budget and build the game of your wildest dreams🔥
I'm learning C, Godot (GDScript), and Game Maker (GML). I've been toying around with the idea of making a simple Catan clone. But now you have me thinking: make the Catan game be the procgen creation of a map where gameplay happens. I mean if you think about it, if someone takes the only spot available that connects stone and brick resources, and you needed those resources, you're going to be mad at that player. Instead of players being mad at each other, this can be used for NPC backstory, detailing why clans rival each other. Obviously would have to make a lot of changes so it's not literally Catan, but it can still be Catan at heart.
I agree, we only just see the very tip of iceberg of what advanced text generation models can do, within next 15 year I can imagine someone would figure out ow to add a "ever expanding dlc" to any story based game were player simply chooses the direction they would wish to go and computer tries its best to create custom designed content on the spot that will never repeat or appear in anybody else game play.
I'm not a game developer but "forever games" has been my ultimate pipe dream and is exactly how you describe it. The closest I have is Minecraft and roguelikes, but these are entirely procedural.
Your content is priceless Timothy!
Can you talk about the music of Arcanum? That really brought the whole game together for me when I played it as a kid. I ended up using it as a sample for some music I made probably a decade later. It just ties everything together and builds so much atmosphere.
Not sure how you'd feel about it but I'd LOVE to see you collaborate with some Fallout modders to either make a new game or an overhaul implementing some of these ideas!
Hope you can pickup that space game idea again!
I love the idea of a system that truly keeps track of things that have an organic place in the same sandbox the player is in.
My favorite example is a bandit camp that raids another one and absorbs all their wealth, and everything in the world kinda just does it's thing like that. Then there's the more literal ecosystem, bears will need to hunt to survive etc. and generally do cool bear things in order to be a bear, not a lifeless npc running off a simple script.
I prefer a shorter game with a handcrafted content rather than a game with a 1000s of bland and samely quests. Even some handmade games can be too long for their own good, there is no need to add some subquality content to the mix. The only way procgen can improve the gaming experience is when it enhances the simulationist aspects, like mentioned in the video factions wars.
Hey Tim, are you sure 'Forever Games' are a good thing for the industry? I think there's a lot to consider in terms of how companies will use it to keep players hooked, and only ever playing their game, or pouring millions of extra dollars into trying to keep a game chugging along when its time in the spotlight is over. There's more I'd say on this, but I'd be opening a massive can of worms.
Oh man I’m so glad you exist
procedural generated AI items, descriptions, even 3D model if you have a 3D game. Infinite unique looking items/armor/weapons. What we need is proceduraly AI generated quests based how you interact with npc's, if you can add voices too and event to that even better. In the future games will be 90% only AI made.
There was a demo from nvidia, showing what AI can do, and was a guy talking ingame through the mic to the npc, and the npc was replying and based on game settings and what you were talking with the npc he could give you clues even missions/quests that was generated just for you
I believe that Handmade and Procedural Content could work well in a cohesive fashion with each other. This is not to portray a full scope, but potentially a method of how Handmade and Procedural Content can coexist with each other. The Handmade content provides the initial structure of the content within an umbrella of rule sets. While the Procedural Content is generated inside of that structure. That way the design can still have some oversight of the generation to some degree and can iterate or balance it because it lives within a defined structure. Perhaps an analogy to convey what I am trying to illustrate is this. For you mmorpg gamers, you know the concept of themepark and sandbox? Think of this more as a sandpark. Something like Everquest. The handmade provides the park portion and the procedural provides the sand portion.
I guess a natural question to ask ourselves is, what kinds of games should be "forever games" and which ones work better without indefinite content?
I find the idea of "forever games" and how they're being thought of rather strange as I was under the impression games you can play "forever" already exist. In fact, the specific games mentioned, Fallout and Skyrim, already hold that distinction, partly due to the modding community's love for them. They've become classics in that sense, but just as the book, Lord of Light ,doesn't change in the least when Tim reads it every year (Tim does), neither do those games. The games themselves remain the same. It is we who keep coming back to them again and again..
You, almost exactly, described how I feel about Starfield right now!
The thing is that as a player as soon as I sniff that a quest is a procedural one (like the Radiant quests) I don't want to spend my time on it. With all the handmade content out there why would I want to spend time playing something randomly arranged by a machine with an unlikely grasp of quality?
The "action game with RPG elements" description didn't come about by accident. "RPG elements" are just that: half-assed and generally unteresting "elements" that lend nothing to gameplay and come across as afterthoughts at best. That's the way supposed "RPGs" in the industry at large have gone and I've not seen any hint of a course correction coming about anytime soon.
Check out Shadows of Doubt, it kinda matches what you described in the video with the procedurally generated NPCs, routines and conflicts/crimes/quests.
I know I'm a little late with this one but I was listening and thinking about it. I feel like the problem with current procedural generation in games is that it makes it all feel disjointed and gamey. Hard to convey the feeling I get from current games trying this. Or at least games with procedural elements that I have tried. My god Tim every idea you had here would neatly solve that disjointed feeling. Would really love to see this
Tim I need you and Leonard to make another isometric fallout with all the procedural content you mentioned
That's a cool ass idea, using the Pokemon style rival system but it's a proc gen player created "antagonist" who builds their own party as the "protagonist" party advances. Twist, you can swap sides at any time and the AI takes over the opposing side using the choices committed (either ai or player made ones) as the seed for continued proc gen. You can play the whole game as 1 party against and AI, or you can take over as frequently or infrequently as you want. Build your own big bad.
If you look at the updatws for Unreal Engine 5 showcase the procdeual generation looks nearly identical to handcrafted content. Procedual generation has the potential to match handcrafting and if it can do that with a massive scale that will change gaming forever.
never ending / xp grind / fetch quests could be really unique and very different each time. Maybe an interesting big story unfolds, while you visit a place you have to steal something from and then you are in the middle of a big fight. There is a llot of potential for such games.
Is there a way to access those GURPS generators today? I'd love to fiddle with them, possibly try to recreate them for coding practice. I watched your videos on those GURPS generators when they came out but I don't remember if there was a way to find them online
Procedural generation seems to be trying to solve another kind of problem: cutting down on the time needed for coding in general rather than making new content. Speeding coding up seems to be the true problem here, and procedural generation at this time is a blunt tool to this end. The procedural problem also highlights the tug of war between quality and quantity in games. Either way, the clearer an idea the developer has for the way a game should be, the less there is the need to make quantity rather than quality.
I'm curious about your thoughts on Starfield's proc gen, especially with your idea for the Universe game
Hi Tim
As you mentioned Barks in this vid I've always wanted to ask...
How come the same bark gets said for different characters using the exact same words even by different actors?
Is it a - we have to pay the writer more if we ask them to make variations - type thing?
Can't you ask the voice over artists to improvise some alternatives in the booth etc? (or does that come with costs too?)
Anyhoo - thanks for all your wonderful stories
:)
Universe, with heavy proc gen, sounds like Elite Dangerous. IIRC David Braben is hugely into proc gen. In honesty, I haven't played that game myself, but I was devoted to Elite as a child.
You've totally sold me on proc gen for the likes of Fallout. I love the idea of the modding community getting their hands on the proc gen engine and coming up with ever more different ways for it to introduce emergent events.
Funny you should mention AI. I was wondering today what your thoughts are about that, so it's good you mentioned you've done a video on that. That's next on the watch list!