I ran across this video many years ago. While helping my wonderful users, I recalled and wanted to show portions of it so many times, because it saves me so much explanation, but I was not able to find the video. I finally found you while helping another person and having another go at finding the video.. Bookmarked, liked and subscribed. ;)
A thought on getting around the issue for the housing meshes. Instead of having a custom mesh for each house, break it up in to modules. For terrain you can move a rock around and scale it, but if you don't throw in at least 2 more boulder types, it's going to become really obvious that every rock is identical. Kind of the same with trees, you can scale and turn one tree mesh, but having 2 or 3 to do it with can help mask what you're doing. So with the houses, just make them modular blocks with their own style. That way you can even do procedural houses that look different from eachother even if a whole town of houses is built off of 6 or so basic house meshes. That way, the most expensive prospect is making a mesh set for culturally distinct locations so that the forest dwelling town can look distinct from a coastal port city by changing out the log cabin texture for a brick one and just stack them more vertically. I know I'm late to the party, but it's a notion.
Sat through the whole video xD, this was very interesting. Definitely appreciating open world games more now for its clever design and workaround in creating a streamlined world that's expansive yet at the same time able to direct its player through the necessary plot points ^_^. Keep up the good work :D!!
This takes me out of games personally. It's been a major problem I've had with fallout games. The illogical, dense and directionless layout of the maps made them disorientating. I feel uncomfortable in the environment and can't learn it. 2 hours exploring a city and for some reason there's only 6 streets I'd loop back on endlessly. But they all lacked spacial queues so I felt lost constantly. That meant that I could never settle into the game. I have 10 hours total in Fallout New Vegas and a grand total of 2 hours in Fallout 3. I settled in Skyrim and I've lost 1600 hours in that game, because it felt like the maps led to places. I got to Whiterun and had an area I managed to learn in the first couple of hours playing there. That got me invested and kept me playing. If I wanted to go from Whiterun to Solitude I'd go between two locations, and the settlements built up somewhat logically. I don't know if it's because it's a medieval setting and the settlements being small didn't need to be hidden, but the thing is I could learn the environment I was in when I was there. I couldn't learn the wilderness. So I enjoyed Skyrim because i could go from pockets of areas filled with characters and life and I could learn the layout of the cities easily. Venture out into the wilderness and familliarity would blend into a confusing maze of pathways in the countryside that felt expansive. Then I'd end up suddenly finding things calm down and I'd be in a new little settlement. So when I design a map I try to have them do that. I create pockets of areas that I try to make sure aren't hiding stuff and trying to confuse you and I try to link them together in places that are confusing. Becuase I find I feel like I want to get out of the confusing areas and I feel immersed in the areas I can learn.
Skyrim was modeled on a functioning society, so everything went sensible places. Fallout 4 has a different priority, and the result is more confusing. Some people have the opposite preference you do, but in general I tend to prefer Skyrim as well.
I agree. The weird-ass scale in a lot of Bethesda games annoys me. Remember some of those 3 house towns in Morrowind? Then you have games like STALKER which have vast areas of nothing, but it's still interesting, and then you hit a true-to-life scale city in Pripyat and Limansk. Granted, you can only explore a few streets, but those few streets feel 100x more real than anything in ANY Bethesda game.
Can you do more of these type of videos Craig? I've watched this ten times already. Love it. I'd love to see you discuss how those city blocks are made more (are sidewalks splines, a bunch of small meshes, one large mesh, etc?). Stuff like that. Thanks!!
Hm! I don't really know if I'm qualified to talk about their tech pipeline in so much detail. But I do plan to make more of these kinds of videos. Just... it's hard to get inspired.
This is a top notch commentary on open world maps. It is well thought out, clearly presented and illustrates so many good ideas for good open-world design. I know I will come back to watch this many times.
Awesome information. I'm just getting back into casual level design myself, and despite this being tips for more professional orientated people or even people making their own games for personal enjoyment, this helps me understand alot more about how I'll go about sculpting the height map and placing meshes and whatnot. Thank you, sir!
I'm an indie coder without a very artistic view, I use a lot of procedural generated buildings with grunge, age & color decal variants that allow it to make them look different when in actuality it's just scheme changes. I went around placing a ton of them only to realize that it was overly cluttered, didn't look good and was a bummer on performance. This explained so so much to me and never really got out and looked above to see all the signs at how they divert you to these designated places. This is awesome thank you so much Craig. I feel like I might actually get somewhere with my terrain design now. Subscribed.
Really great talk man. This was all stuff I feel like I know on some level but you just lay everything out so clearly and consciously, as usual. I couldn't help but think about playing Morrowind with mods that remove the fog and add distant lods...you can really feel the lack of these things and it makes the game feel really strange
I absolutely love this video, you do an amazing job of explaining this, definitely deserve all of the likes and subs and views you get, keep up the good work! :)
19:20 could always have modular buildings and randomize them a bit, I'd make a few packs though to add more variance. Really great exploration into open world map development, you were thorough and precise in your explanations and I enjoyed it a lot thanks! :D
I remembered when I played Zelda BOW and the game lets you free to go anywhere but at the same time it lets dangerous enemies out of the track so you get more confident to go where they want instead of facing Dark souls bosses everywhere.
I love things like this, thanks. I'm working of a game of my own in unreal 4 right now and I really wasn't sure about the whole using meshes for the rocky parts of hills and mountains thing. looks like I'll be reworking a lot of my map. but I really love thinking about the theory behind intuitive and clever level design. I just recently watch a video about how the level design in dark Souls fits in with it's themes and motifs.
Awesome video and explanation of how to mask the hightmap. I can't tell you how many times I've created a world and decided that it was way to hard to make it feel like a real place. With you dissecting fallout 4 I have a better understanding of how the world is made. I only dabble in level design for fun. Can't wait to try some of these tricks.
Very nicely and professionally covered! Really fun to watch and i do agree with the whole sign post technique. Particularly the one at - 23:08 - it totally reminded me what ID Software did in their game " RAGE " where they used a huge highway which was destroyed on the end of the road clearly indicating the end of line as there was an enormous drop below leading to a massive canyon, basically saying " End of a trip, this is where it ends, but you can go on if you dare, but look, heres a dungeon you can go in " boom, instant eye candy to see the hills in background with wonderful skybox ( one of the best i seen in that game i have to be honest, most likely due to the fact they used those mega textures ) and it really does make player feel its huge. Many people were dissapointed in Rage cause they felt it was too small, but it actually is very well done in a way where it does not necessarily prevent you to explore more, but as you said - when you go where the content is - you feel very satisfied. Also a good thing to remember! They gave players buggies to drive around as the paths were really long but guess what - You are free to go by foot... so that is a huge freedom choice - either take it and walk it all which takes you way too much time, or drive it - but the choice is there for you to decide upon since some people might prefer to run around while looking at stuff rather than speed through it and that is really what makes difference in open world games indeed. I remember to this day the part where you enter the city with huge skyscrapers which by the way were a massive sign posts showing off proudly over hills which told you " You will go here one day, but not yet " And my god... when i was there, even tho it was pretty straight strict path - it felt so natural with all the obstacles, events and eye candies all around, the massive buildings, demolished cars, atmosphere sounds, it felt real and this is what is really important about making it look realistic ( but not actually real ) because when it would be too real, it would lose its magic, since you would need to start putting way too many limits that we are used to in real life and that is not fun. Sometimes crazy stuff or logic adds to the fun, its games after all : )
When you say using different meshes for each building is expensive, do you mean in production/creation, rendering, or memory? What about highly modular buildings?
Although I know most of the concept you talk about, it was a very insightful video! As you know I'm obsess with PCG and bringing it to the next level, one requirement is to break down conceptually elements much more than with hand craft building, the computer won't magically solved ambiguity with intuition like human do. A lot of the burden is, therefore, to translate knowledge into "conceptual folder or alias" for the machine to manipulate, and this video made me rethink one or two approach into implementing these concepts MUCH BETTER! 1. sightmark: Instead of organizing open world generation with landmark (big far away signpost) and POI (small signpost, eventually activity point, those distinction aren't very useful for a human though, it's more for the computer code), it's best to organize them in sightmark, ie point where these elements are visible. It's much better for organization because it has implicit quality to manipulate flow, it effectively encode edges instead of point like landmark and POI, and edges are much better at conceptualizing flow. It bridge path with sight and aesthetics much more fluently for an algorithm, it's better to make code that reason on both at the same time! 2. block prompt A form of visual blocking that actually suggest more, like path corner. Small concept, but the devil is in the details. 3. soft blocker Well this one is obvious, I know it, I use it, but for some reason, I have a notepad full of note and I forget to note this one, shameful of me, I'm embarrassed, that's why you don't want intuition when doing level design. You want explicit not implicit. A last and useless observation, more like a small unfounded rant, forgive me! One thing that unnecessarily bother me when looking at level generation is how people tend to rely on one single magic technique, when they should be combine together. You don't always need a voxel terrain with .25 precision, and you don't have to only have that, heightmap are fine too most of the time! When you showed the river, I had an epiphany, this mesh could be generated by a voxel box spawned and deformed along the edge of the river. In fact I think 2m resolution size voxel terrain is generally mostly fine to give the overall shape of terrain which can be decorated with tesselated perturbed surface, and additional details with sparser mesh generated by voxel box freely span on the terrain. That's how we reason with handcraft generally. 2m is just the smallest size of a room with non perturbed shape (can contain the character, or the size of the smallest road with enough movement), it can LOD much better (it's low frequency and the hi frequency surface perturbation can be generated later at less cost) and is actually better at conveying gameplay metric than fine resolution, therefore much better for traversal. It also prevents the generation from doing crazy difficult to traverse hi frequency litter (or to get rid of it) and is 32760 times less dense in data than .25 resolution, which solve so many problem I don't even know where to start (like generation time and memory problem).
Structurally, voxels might be useful here, but in practical terms one of the advantages of meshes is how high-quality they are. You'd have to have a really great voxel engine to have realistic river berms. As to sightmarks, I don't think I agree, because they require you to always know where the player is in order to calculate where the player can see. By breaking it down into short and long signposts, you can use a simple heuristic that works out regardless of the direction the player is viewing from. Other than that, I think I agree with your points.
Sightmark: I don't agree, they are supposed to be relation, ie they aren't "point" (maybe in the code but only to organize space around them), but they are the relation between landmark, blocker, path, etc ... Knowing where the player is, that's easy anyway, as easy as in handcraft, you might never know exactly, but you can prescribe place (push and pull logic, also because level is made of traversable and non traversable space), especially if you place them along path. The thing is that to allow logical sight breadcrumbing like the one you mention with the tower that is then hidden to redirect to another landmark. With only landmark and POI you only have attractor to organize space, breadcrumbing go from attractor to attractor and don't allow logic like temporary landmark (or path logic) that are only relevant up to some distance, then blocked at another (which would be two sightmark organized logically using a design pattern). That said you can be a landmark and a sight mark at the same time (see ubisoft tower). Voxel: I don't mean pure voxel anyway, for me voxel is almost always about creating a supporting volume to decorate in some way or another, mostly using surface perturbation. So it's easy to conceive high detail mesh with low detail voxel support. Hand craft mesh will always be repetitive, voxel is just a tool for procedural mesh, and a procedural mesh demand as mush care to handcraft (per type, it wouldn't actually save time at all) than a hand made mesh anyway, the only advantage is that it wouldn't repeat. I think it's doable to make to hi quality procedural mesh mixing LOCAL freeform voxel objects (ie can be deformed and transformed freely) and other technique. It wouldn't be a huge voxel terrain like we are used to, just object with just in time generation inside their own bound. Local voxel for details allow to avoid the huge cost of full voxel terrain *random rambling ahead* HOWEVER, a just in time voxel generation that only spit out the mesh and don't keep the data can be an alternative I'm just thinking of now, I mean you don't need the data once teh mesh is generated because it's all aligned at integer distance, therefore detecting a mesh inside a specific range would allow to do the necessary modification implicitly by checking neighbourg (and you could hide extra data in unused vertex data). Mesh being supposed to be enclosing, inside empty space would only be very specific failure case. Data being procedural, they can just be queried at running when necessary (when modifying a black and carving inside the "empty space"), it cost just a bit more than the typical look up, but they only cost a lot when filling an entire cubic chunk, simple query should balance the memory/processing cost, also, most of the time, manipulation is done at "surface voxel" not interior one. And if you really need the look up, using bit array greatly reduce memory cost from 32^3 (32mo) to 32^2 (1mo) byte, that's order of magnitude less, while it wouldn't store enough data, you can still generate them using procedural query just in time and skip the data when unnecessary to keep process small. It would also speed generation time in a format like minecraft, where there is solid and empty voxel, things like ore being generated as internal (data being mask out) you wouldn't generated them at all until they are revealed (ie become surface objects) which mean you can separate the carving generation (solid and air) from specific voxel, and move the generation of specific voxel in a just in time generation, when block are shown in screen or carved, with just in time procedural query. therefore unnecessary unseen data isn't generated at all until it's pertinent. Another trick would be path find from the camera position's chunk, to generate each chunk using a kind of frustrum filling djikstra by view connection to each successive chunk, so that chunk not seen aren't generated, given the cubic cost of generating voxel (memory and process), this overhead might help culling a lot of unseen data. I'm trying to determine an early skip for chunk if empty (maybe a two step generation, at chunk frequency then at voxel frequency), but I'm blanking. I'm sure I hadn't think this through lol just typing down to keep note before forgetting it. Sorry to use the comment as a scratch pad for sudden idea!
Which one are you talking about lol If it's teh local voxel, they are are design for just that. The rambling is about more general idea that pop up and have nothing to do with the idea of local voxel mesh. But yeah I'm thinking about it, I need to start programming them, but I need to implement half edge and convex hull before that, and there are not related except they are part of the same project (spore ish open world creature phase).
I just wanted to point out that the reason that Bethesda games don't use simple textures for roads and instead choose to utilize meshes doesn't necessarily have to do with game design. The Gamebryo engine simply isn't capable of texture rotation, and as a result paths and road textures can look jagged and strange when misaligned with the height map's vertices. You notice this a lot in Skyrim's many dirt paths. With a road's smooth, bending lines and edges, the only way to get them in-game is to kind of cheat and connect two road models which are cracked at the ends. On the other hand, this does result in some very nice and rugged looking roads. Nice video overall!
Modular packs with damaged piece alternative pieces so every building can be any size and have damage on any piece. I just downloaded a pack like this actually
Really interesting stuff. I didn't really know how huge signposting was, but now I'm really thinking about it. It sounds like you're not a huge fan of proc-gen open worlds, but one of the takeaways I have here is that, if one is committed to procedurally generating an open world, then they should be generating paths and the core of their generator should be a lot more like a procedurally generated dungeon, just without the roof than any kind of statistical simulation of a world. At least, if the world is supposed to be as interesting as the AAA open worlds.
Every tool has its use. I'm not against generated worlds, but I found a lot of people making fairly basic mistakes with handmade worlds. Their strength is in things generated worlds haven't got figured out yet.
I agree with Perko, I'm big on PCG world but most of my focus has been about to solve basic design problem rather than automation problem, if you don't have solid design understanding, your algo will not do the work for you. As a culture, pcg people are still mire into "magic math" perception of pcg with the occasional design "hack" like generating random critical path, but it's not there where they basically implement actual rules of design as generator. The curse of math pcg is that they only mimick the structure, they don't think about functional meaning of the structure or think about the structure of intent. So it's like having a sentence generator and all phrase are like "colorless green sleeps furiously", it's grammatically correct but mean nothing. That's why you have so much focus on making thing look good and constrain it to be playable rather than algo that generate actual design goal to fulfill. In fact the way algo works attract more attention than what they actually do and why. The poster child of this mentality is the article about tiny keep's random room, that just pile room on top of each other, use physics code to separate them, then construct a graph of room that connect to each other, then remove some line in cycle to have loop and dead end, that's the organisation equivalent of a spaghetti plate, why a room is there, why it has this size, how does the level flow and why, those are clearly not in the algo. It doesn't create meaning, it's only accidental. BUT THE PROCESS IS PRETTY TO LOOK AT! Algo should be implementation of solid understanding of game level design. It also mean that you should have solid understanding, beyond simple intuition, of level design to begin with, something even in the game industry is a bit lacking. Man! everytime I'm getting triggered by PCG discussion lol
Thanks for the clarification Craig, I appreciate that. Neoshaman, I actually agree. That's the direction I think I was going talking about dungeon design instead of just dotting a world, but it looks like you're a few steps ahead of me. Yeah, you need grammar, you need a story, themeing, you need sightlines as part of your design, all stuff that hasn't been done yet in a really admirable fashion with algorithms. It will, of course, in time, but as you said, it requires an understanding that you can actually communicate in code. But that all comes back to Points of Interest or SignPosts, and these are things that are fundamentally unique, where as PCG is fundamentally not. So, that's a thing. Unless, of course, you're doing literal signposts, because y'know, they're all the same.
"But that all comes back to Points of Interest or SignPosts, and these are things that are fundamentally unique, where as PCG is fundamentally not. " I totally disagree about that, PCG has some fundamental limit but the reason it's popular is to make things unique. POI and signpost are like the easiest thing to generate. The real problem is what I call te dictionary problem, which is multi faceted. At a simple level it's the bowl of owkmeal problem, all bowl of oakmeal are different, but still are essentially the same thing. The next step is the infinite chair problem, even though you can generate all chair in existence, it's still just a chair, it as limited function and expression. The next step is the expression problem, even though you have more than chair, you have to put them together in a way that makes sense and express something, you can have infinite room arrangement but only some have meaning, like the office rooms, the funeral room, the living room, etc... The problem here is that even if your generator is expressive, this expression comes from human experiences, that's unique pattern to code, which comes back full circle to handcrafting things, because you need to encode specificity, human don't have necessary access to other human experience, I may know what a gallodrome looks like in martinique but not in america. So the process of filling a pcg algo is basically encode by hand those specificity. Even though they are contain implicitly in the possibility space, and the algo can generate something we can recognize as part of the human experience, it's only accidental and would not be controlled. Since basically those element are description, they are like entry in a dictionary, hence the name. A large dictionary mean that you have good vocabulary to be expressive. Expressivity would be the fitting range of an algorithm that match human experience, ie its ability to compose new description that make sense to human experience. So the irony is the opposite of what you described, you want pcg to be less "unique" in the absolute sense, as absolutly unique thing wouldn't be recognizable as part of the human experience. Hence why meaning and semantic must generated and manipulated to generate fitting composition. In another comment I talk about how I plan to create an algo that would make sense of sightseeing composition within pcg. Hence why I compare pcg to a linguistic process rather than a mathematical one. Also hierarchical generation is a quick and easy way to encode flexible meaning in a pcg algo. The highest composition capture meaning and intents as structure and propagate to lower structure. This mean that it's easier to generate using correlation rather than causation. Sightseeing and signposting would be notoriously hard if the terrain is generate and the right need to be found ... it's best to design the sight first and then generate the terrain to fit it. The correlation problem can be easily illustrated with river generation, it's notoriously hard to find adequate river with enough control (for gameplay, sight seeing, etc ..) for a given terrain, but the terrain cause the river and it can lead to non predictable or unnatural composition if forced. We can reverse cause to correlation, since a river always flow down, it's way easier to fit a mountain to a river rather than the contrary, the slope and direction of the river is the same as the mountain, which allow better control as a river is easier to generate. Applying this to everything allow to generate complex phenomena, like npc interactions and history, without the simulation hassle we use to have. That's the basis of the infinite persistent NPC generation I'm planning to do.
That's pretty doggone fascinating. I deeply appreciate the linguistic angle, and I see what you mean about absolutely unique. I'm not sure how signposting could be the easiest thing in the whole process, since as I understood from this video, major signposts are essentially landmarks, and I'm a little iffy on a relatively trivial landmark generator. As far as grammar and structure and rules, one thing that comes to mind is blockbuster movies. These follow a very solid and predictable three act structure. They also have a limited number of 'types' or genres. People continue to go back to them, however because of the large 'dictionary' and the potential to find rare 'uniques' which may not be truly unique, but are unique to their experience because of how rare they are, and yes, how common they are to the human experience. That kind of dictionary, color palette theming, juxtaposition, and a hundred other tools of sound, visuals and possibly even mechanics would be quite an achievement. We saw what it took for the No Man's Sky team just to create a visual grammar that was appealing. I'd love to work on the kind of algo you're talking about... but I have to admit, it would be a beast like no other.
Spyro or Crash. You can have an open world to explore but it will have less stuff in it and is mostly just flat or you can have smaller linear maps with maybe a little more up and down slopes but are filled with more objects and living things. Compare Rayman 3 to Silent Hill 2 or Perfect World. In Rayman 3 you can see from the introduction and first level that it's rich with windows and therefor homes built into the massive trees, garden gnomes and creatures littered around and small bugs flying around along with mushrooms and random holes with hills going up and down and bridges and hanging platforms with detailed background models and objects have small details added like lights. Then in Silent Hill it's mostly flat empty streets with every little in them as well as each building being a separate level with the fog being used to cover the low capabilities. Perfect World is an mmo and like any mmo it has a lot of open areas with maybe just large trees and empty roads with specific areas highlighted with npcs and other objects
Not a game maker (not got a head for making code) but I found this very interesting. Also helpful for when it comes to making mods for Bethesda games :)
OblivionSilk You actually barely need to know code to create a game. Of course it helps, but creating a game is often just a bit more complicated than creating a game :)
This was a very information dense video that you've made flow quite nicely by using Bethesda's signpost pathing. Is there a good guideline of the variety of content that can be found at the end of the signpost rainbow? I'm applying a lot of this information to my exploration driven Dungeons & Dragons game and following the principle of breaking up patterns (like your point on the dike) it seems that I should probably consider other things to place at my signposts other than "empty, monster, special, trick, trap or treasure".
For a tabletop game? If you know your players, figure out what things they might like that the other players might be a bit disinterested in, then add that as spice on top of a more fundamental reward. For example, finding a magic rune in a goblin's pocket, or a diagram, or a widget, or a friendly talking squirrel, or a strange pipe begging to be smoked. By having the right player find them first, you can prevent the other players from putting the brake on. Just be sure that you fold it back into the story with other finds: you want to keep the right number of loose ends. Too many and things follow apart. In terms of physical signposts, it's mostly about the flavor you give when you describe it.
That's pretty solid advice, especially concerning loose ends with more story-focused campaigns. One of the things I find about location design is that it can be difficult to actually give them satisfying longevity in play. Most of the time you are "clearing" content and after the first couple of times you deal with signposts, they're only really useful as navigational markers.
I don't quite understand how developers are able to work on the same asset of map simultaneously. What kind of procedures or pipelines are used so that you can have multiple people editing a single file at the same time without any of the modifications contradicting one another or corrupting?
This is super interesting and informative, thank you! Unifying the meshes and height map would reduce the vertices, but it would make texturing a nightmare, right?
How to get around needing tons of building meshes for cities: MODULARITY I'm using modular meshes with procedurally generated buildings for my game that takes place in an abandoned city. The buildings have randomized numbers of floors, facades, side-brick work, windows, chances for roof accessories, awnings, air conditioners, decals for damage/ graffiti... not hard to set up, but it can get a little tedious. I bit the bullet and bought a modular building pack with 5 types of buildings with mix and match ground floors, middle floors, and top floors, but most of the big world meshes are my own (roads/ sidewalks/ building greebles like pipes and vents).
@@CraigPerko Thx for the reply, but I have to disagree. When you are making basically 4k openworld, you would want to be really careful about the poly batches you are loading at one time. Maybe one day when we have super computer it won't, but that is decades to come.
That's a complicated situation because not only are things moving rapidly, but you probably can't block their sight. I'd focus a lot more on having long, medium, and short-range POI.
All of these things depend on your project. If you're just getting started, use a small map: the entirety of Fallout 4 is small enough that you'll never run into floating point errors. Smallish height maps are fine - 1024x1024 or 2048x2048 should work fine, because that can be a kilometer across without any problem. If you need larger or denser spaces, you should use multiple height maps and only load the ones you need or your performance will suffer. As for buildings: they need to be tall enough to do their job. Whether that job is acting as a soft barrier to let you split one side of the building from the other, or maybe that job is acting as a tower to climb, or as an instance gateway to a dungeon, or as a standalone point of interest to sack... design the building to do the job you need it to do.
Thanx but what I don't understand is how many pixels I stretch the 1024x1024 map. how big should it be in world space floating points? It cannot really be 1024x1024 because my screen is bigger than that so what size is it really? for example if I am using a triangle strip with the 1024x1024 height map - how big should each triangle be to represent a kilometer?
You don't have to build the terrain mesh yourself, but typically a pixel is either one meter or half a meter. I recommend using Unity's built-in terrain, or forking out for something like Gaia if you can afford it.
I am building the mesh myself but I did some googling and figured out that the "scale factor" is what I was asking about and that its up to the programmer to pick a scale factor for their world.
They have an aggressive LOD system that uses occlusion. I think it also does predictive loading, which is why my high-speed, high-altitude antics caused loading pauses. Unity can pull the same basic tricks, they're not unusual.
I see, thanks. I've played around with LOD in Unity a while ago but I couldn't find a way to make it unnoticeable enough. Maybe I should look into it again..
You absolutely need LOD to do things like this. I generally use a simple decimated version of the mesh, it's usually pretty hard to notice if you decimate it the right amount.
I gave it another go and it came out better this time. I think the quality of the meshes is crucial... I tried the mixamo decimator which seems to work quite good for organic stuff. Also the Blender decimate modifier can exclude a set of verts which is nice :) Now I'm thinking about ways to integrate LOD level generation in my modeling process for the more hard surface models. Do you have any workflow tips you would share?
The easiest way to do it in Unity/Blender is to simply use linked duplicates and add a carefully-tweaked decimate modifier to each. Named correctly, Unity will automatically put them in an LOD object, and because they're linked, modifications will propagate. Hard surfaces do well with planar decimation.
No. Bethesda did. The flip is that you can. It just takes a lot of time. especially making buildings with working doors texturing, baking etc. Very few people are taking worldcreation as seriously as the man in this video but they deffinately should. You've gotta be very consious while doing these things. I do so myself and have found that the difference is night and day. It also takes a lot of time to really understand the programs you use but it is worth it if you''re a creative person.
my wife was killed...long ago...my son was kidnapped...long ago...when waked up the second time I was so angry, I jumped off the cliff...searching for a sniper and forgot my family
I was taking a look at GTA V world using CodeWalker, it pratically doesn't have terrains, at least heightmap terrains... It's all meshes, you can turn on the wireframe and you will see that entire topology of the models comes together in a very harmonic way, roads, terrains and buildings, like they were modelled together, it's mind blowing! They probably has some very different workflows and ways of doing things that unfortunately they never disclose... :( Very nice video by the way, it helps a lot and it brightens a lot of questions I had regarding how fallout 4 world was designed. Thanks!
By the way nice vid! I clicked this vid to check out the map, and I noticed it was just a clip to learn about making games. Before I pressed the back button, I started to get interested, so I sat back and started watching ;)
Can't say I disagree since I'm only one person, but my experience of signposting and blocking in fallout 4 is very different to what you portray and what Bethesda intended, at least after Concord, the feel of almost being railroaded right to Diamond City, the feel of needing to shun any landmarks because they're so limiting to ones movement (especially in combat), the feel of almost always being told what a site is from miles off. As said I don't think I'm in the majority nor even intended player base, but for me, it just makes me stick to the wilderness they want me to avoid unless I'm being railroaded to the next quest.
Fallout 4 had some serious weaknesses, and the biggest one for me was the HUD icons that encouraged you to not explore. That said, I didn't feel any of the other things you seem to have felt - exactly the opposite. I didn't even get to Diamond City until ~15 hours into the game. Some setpieces were very rigidly laid out, but the world as a whole didn't feel that way to me. I'd be interested to know more about how you felt, because it's clearly different enough that it should be taken into account.
FYI, I would really watch the hell out of a series like this, flying around game maps explaining design tricks and techniques.
Well, that's not a bad idea.
yes pls, have fun with it
+Craig Perko DO IT....yesterday you said tomorrow...Nothing is impossible..JUST DO IT.
I would especially appreciate an episode about the Breath of the Wild (at least part of it).
I don't own a Switch.
Wow this is really interesting, I would watch a whole series on this type of stuff
I ran across this video many years ago. While helping my wonderful users, I recalled and wanted to show portions of it so many times, because it saves me so much explanation, but I was not able to find the video. I finally found you while helping another person and having another go at finding the video.. Bookmarked, liked and subscribed. ;)
Welcome back, hah!
A thought on getting around the issue for the housing meshes. Instead of having a custom mesh for each house, break it up in to modules. For terrain you can move a rock around and scale it, but if you don't throw in at least 2 more boulder types, it's going to become really obvious that every rock is identical. Kind of the same with trees, you can scale and turn one tree mesh, but having 2 or 3 to do it with can help mask what you're doing. So with the houses, just make them modular blocks with their own style. That way you can even do procedural houses that look different from eachother even if a whole town of houses is built off of 6 or so basic house meshes. That way, the most expensive prospect is making a mesh set for culturally distinct locations so that the forest dwelling town can look distinct from a coastal port city by changing out the log cabin texture for a brick one and just stack them more vertically. I know I'm late to the party, but it's a notion.
Sat through the whole video xD, this was very interesting. Definitely appreciating open world games more now for its clever design and workaround in creating a streamlined world that's expansive yet at the same time able to direct its player through the necessary plot points ^_^. Keep up the good work :D!!
This takes me out of games personally. It's been a major problem I've had with fallout games. The illogical, dense and directionless layout of the maps made them disorientating. I feel uncomfortable in the environment and can't learn it. 2 hours exploring a city and for some reason there's only 6 streets I'd loop back on endlessly. But they all lacked spacial queues so I felt lost constantly. That meant that I could never settle into the game. I have 10 hours total in Fallout New Vegas and a grand total of 2 hours in Fallout 3.
I settled in Skyrim and I've lost 1600 hours in that game, because it felt like the maps led to places. I got to Whiterun and had an area I managed to learn in the first couple of hours playing there. That got me invested and kept me playing. If I wanted to go from Whiterun to Solitude I'd go between two locations, and the settlements built up somewhat logically. I don't know if it's because it's a medieval setting and the settlements being small didn't need to be hidden, but the thing is I could learn the environment I was in when I was there. I couldn't learn the wilderness.
So I enjoyed Skyrim because i could go from pockets of areas filled with characters and life and I could learn the layout of the cities easily. Venture out into the wilderness and familliarity would blend into a confusing maze of pathways in the countryside that felt expansive. Then I'd end up suddenly finding things calm down and I'd be in a new little settlement.
So when I design a map I try to have them do that. I create pockets of areas that I try to make sure aren't hiding stuff and trying to confuse you and I try to link them together in places that are confusing. Becuase I find I feel like I want to get out of the confusing areas and I feel immersed in the areas I can learn.
Skyrim was modeled on a functioning society, so everything went sensible places. Fallout 4 has a different priority, and the result is more confusing. Some people have the opposite preference you do, but in general I tend to prefer Skyrim as well.
I agree. The weird-ass scale in a lot of Bethesda games annoys me. Remember some of those 3 house towns in Morrowind? Then you have games like STALKER which have vast areas of nothing, but it's still interesting, and then you hit a true-to-life scale city in Pripyat and Limansk. Granted, you can only explore a few streets, but those few streets feel 100x more real than anything in ANY Bethesda game.
you are plagued by needing linear,on rails ,boring games
Great vid, enjoyed it throughout. Especially liked the clear narrative and the pauses really helped to get the message.
Really well done Craig. Thank you for these observations.
This is one of the most informative videos on the subject I've ever seen.
Can you do more of these type of videos Craig? I've watched this ten times already. Love it. I'd love to see you discuss how those city blocks are made more (are sidewalks splines, a bunch of small meshes, one large mesh, etc?). Stuff like that. Thanks!!
Hm! I don't really know if I'm qualified to talk about their tech pipeline in so much detail. But I do plan to make more of these kinds of videos. Just... it's hard to get inspired.
I never realized how much thought and intricate detail goes into designing maps and directing the player!
This is a top notch commentary on open world maps. It is well thought out, clearly presented and illustrates so many good ideas for good open-world design. I know I will come back to watch this many times.
Awesome information. I'm just getting back into casual level design myself, and despite this being tips for more professional orientated people or even people making their own games for personal enjoyment, this helps me understand alot more about how I'll go about sculpting the height map and placing meshes and whatnot. Thank you, sir!
I'm an indie coder without a very artistic view, I use a lot of procedural generated buildings with grunge, age & color decal variants that allow it to make them look different when in actuality it's just scheme changes. I went around placing a ton of them only to realize that it was overly cluttered, didn't look good and was a bummer on performance. This explained so so much to me and never really got out and looked above to see all the signs at how they divert you to these designated places. This is awesome thank you so much Craig. I feel like I might actually get somewhere with my terrain design now. Subscribed.
Really great talk man. This was all stuff I feel like I know on some level but you just lay everything out so clearly and consciously, as usual.
I couldn't help but think about playing Morrowind with mods that remove the fog and add distant lods...you can really feel the lack of these things and it makes the game feel really strange
Yeah, one of the biggest tricks in open world games is controlling sightlines. Morrowind did it with fog, which is considered oldschool these days.
i just got .03% smarter.... time to unlock my new science perk XD
I absolutely love this video, you do an amazing job of explaining this, definitely deserve all of the likes and subs and views you get, keep up the good work! :)
19:20 could always have modular buildings and randomize them a bit, I'd make a few packs though to add more variance. Really great exploration into open world map development, you were thorough and precise in your explanations and I enjoyed it a lot thanks! :D
Great video. Made the gears turn quite a bit and they still are after it ended. Thanks for the food for thought
I remembered when I played Zelda BOW and the game lets you free to go anywhere but at the same time it lets dangerous enemies out of the track so you get more confident to go where they want instead of facing Dark souls bosses everywhere.
I love things like this, thanks. I'm working of a game of my own in unreal 4 right now and I really wasn't sure about the whole using meshes for the rocky parts of hills and mountains thing. looks like I'll be reworking a lot of my map. but I really love thinking about the theory behind intuitive and clever level design. I just recently watch a video about how the level design in dark Souls fits in with it's themes and motifs.
I just found your channel and you make super interesting videos. Keep it up!
Really interesting information Thanks for breaking this down so well
19:07 The roads here in the Netherlands are this size.
Well, not all of them, but a big part of our roads.
One question, what is so special about a indie game?
I never got why it was so special, haha.
Verry intresting video!
Just that there would be a lot more of these games, with a wider variety of sensibilities.
Craig Perko Thnx. :)
Cuz bicycles need less space than cars, lol)
Awesome video and explanation of how to mask the hightmap. I can't tell you how many times I've created a world and decided that it was way to hard to make it feel like a real place. With you dissecting fallout 4 I have a better understanding of how the world is made. I only dabble in level design for fun. Can't wait to try some of these tricks.
Very nicely and professionally covered! Really fun to watch and i do agree with the whole sign post technique.
Particularly the one at - 23:08 - it totally reminded me what ID Software did in their game " RAGE " where they used a huge highway which was destroyed on the end of the road clearly indicating the end of line as there was an enormous drop below leading to a massive canyon, basically saying " End of a trip, this is where it ends, but you can go on if you dare, but look, heres a dungeon you can go in " boom, instant eye candy to see the hills in background with wonderful skybox ( one of the best i seen in that game i have to be honest, most likely due to the fact they used those mega textures ) and it really does make player feel its huge.
Many people were dissapointed in Rage cause they felt it was too small, but it actually is very well done in a way where it does not necessarily prevent you to explore more, but as you said - when you go where the content is - you feel very satisfied.
Also a good thing to remember! They gave players buggies to drive around as the paths were really long but guess what - You are free to go by foot... so that is a huge freedom choice - either take it and walk it all which takes you way too much time, or drive it - but the choice is there for you to decide upon since some people might prefer to run around while looking at stuff rather than speed through it and that is really what makes difference in open world games indeed.
I remember to this day the part where you enter the city with huge skyscrapers which by the way were a massive sign posts showing off proudly over hills which told you " You will go here one day, but not yet "
And my god... when i was there, even tho it was pretty straight strict path - it felt so natural with all the obstacles, events and eye candies all around, the massive buildings, demolished cars, atmosphere sounds, it felt real and this is what is really important about making it look realistic ( but not actually real ) because when it would be too real, it would lose its magic, since you would need to start putting way too many limits that we are used to in real life and that is not fun. Sometimes crazy stuff or logic adds to the fun, its games after all : )
Oh, I haven't played Rage in a long time, but I think I remember that!
Great video! Super informative, it's like composing your world for the player.
When you say using different meshes for each building is expensive, do you mean in production/creation, rendering, or memory? What about highly modular buildings?
Although I know most of the concept you talk about, it was a very insightful video!
As you know I'm obsess with PCG and bringing it to the next level, one requirement is to break down conceptually elements much more than with hand craft building, the computer won't magically solved ambiguity with intuition like human do. A lot of the burden is, therefore, to translate knowledge into "conceptual folder or alias" for the machine to manipulate, and this video made me rethink one or two approach into implementing these concepts MUCH BETTER!
1. sightmark:
Instead of organizing open world generation with landmark (big far away signpost) and POI (small signpost, eventually activity point, those distinction aren't very useful for a human though, it's more for the computer code), it's best to organize them in sightmark, ie point where these elements are visible. It's much better for organization because it has implicit quality to manipulate flow, it effectively encode edges instead of point like landmark and POI, and edges are much better at conceptualizing flow. It bridge path with sight and aesthetics much more fluently for an algorithm, it's better to make code that reason on both at the same time!
2. block prompt
A form of visual blocking that actually suggest more, like path corner. Small concept, but the devil is in the details.
3. soft blocker
Well this one is obvious, I know it, I use it, but for some reason, I have a notepad full of note and I forget to note this one, shameful of me, I'm embarrassed, that's why you don't want intuition when doing level design. You want explicit not implicit.
A last and useless observation, more like a small unfounded rant, forgive me!
One thing that unnecessarily bother me when looking at level generation is how people tend to rely on one single magic technique, when they should be combine together. You don't always need a voxel terrain with .25 precision, and you don't have to only have that, heightmap are fine too most of the time!
When you showed the river, I had an epiphany, this mesh could be generated by a voxel box spawned and deformed along the edge of the river. In fact I think 2m resolution size voxel terrain is generally mostly fine to give the overall shape of terrain which can be decorated with tesselated perturbed surface, and additional details with sparser mesh generated by voxel box freely span on the terrain. That's how we reason with handcraft generally.
2m is just the smallest size of a room with non perturbed shape (can contain the character, or the size of the smallest road with enough movement), it can LOD much better (it's low frequency and the hi frequency surface perturbation can be generated later at less cost) and is actually better at conveying gameplay metric than fine resolution, therefore much better for traversal. It also prevents the generation from doing crazy difficult to traverse hi frequency litter (or to get rid of it) and is 32760 times less dense in data than .25 resolution, which solve so many problem I don't even know where to start (like generation time and memory problem).
Structurally, voxels might be useful here, but in practical terms one of the advantages of meshes is how high-quality they are. You'd have to have a really great voxel engine to have realistic river berms.
As to sightmarks, I don't think I agree, because they require you to always know where the player is in order to calculate where the player can see. By breaking it down into short and long signposts, you can use a simple heuristic that works out regardless of the direction the player is viewing from.
Other than that, I think I agree with your points.
Sightmark:
I don't agree, they are supposed to be relation, ie they aren't "point" (maybe in the code but only to organize space around them), but they are the relation between landmark, blocker, path, etc ... Knowing where the player is, that's easy anyway, as easy as in handcraft, you might never know exactly, but you can prescribe place (push and pull logic, also because level is made of traversable and non traversable space), especially if you place them along path. The thing is that to allow logical sight breadcrumbing like the one you mention with the tower that is then hidden to redirect to another landmark. With only landmark and POI you only have attractor to organize space, breadcrumbing go from attractor to attractor and don't allow logic like temporary landmark (or path logic) that are only relevant up to some distance, then blocked at another (which would be two sightmark organized logically using a design pattern). That said you can be a landmark and a sight mark at the same time (see ubisoft tower).
Voxel:
I don't mean pure voxel anyway, for me voxel is almost always about creating a supporting volume to decorate in some way or another, mostly using surface perturbation. So it's easy to conceive high detail mesh with low detail voxel support. Hand craft mesh will always be repetitive, voxel is just a tool for procedural mesh, and a procedural mesh demand as mush care to handcraft (per type, it wouldn't actually save time at all) than a hand made mesh anyway, the only advantage is that it wouldn't repeat. I think it's doable to make to hi quality procedural mesh mixing LOCAL freeform voxel objects (ie can be deformed and transformed freely) and other technique. It wouldn't be a huge voxel terrain like we are used to, just object with just in time generation inside their own bound. Local voxel for details allow to avoid the huge cost of full voxel terrain
*random rambling ahead*
HOWEVER, a just in time voxel generation that only spit out the mesh and don't keep the data can be an alternative I'm just thinking of now, I mean you don't need the data once teh mesh is generated because it's all aligned at integer distance, therefore detecting a mesh inside a specific range would allow to do the necessary modification implicitly by checking neighbourg (and you could hide extra data in unused vertex data). Mesh being supposed to be enclosing, inside empty space would only be very specific failure case.
Data being procedural, they can just be queried at running when necessary (when modifying a black and carving inside the "empty space"), it cost just a bit more than the typical look up, but they only cost a lot when filling an entire cubic chunk, simple query should balance the memory/processing cost, also, most of the time, manipulation is done at "surface voxel" not interior one.
And if you really need the look up, using bit array greatly reduce memory cost from 32^3 (32mo) to 32^2 (1mo) byte, that's order of magnitude less, while it wouldn't store enough data, you can still generate them using procedural query just in time and skip the data when unnecessary to keep process small.
It would also speed generation time in a format like minecraft, where there is solid and empty voxel, things like ore being generated as internal (data being mask out) you wouldn't generated them at all until they are revealed (ie become surface objects) which mean you can separate the carving generation (solid and air) from specific voxel, and move the generation of specific voxel in a just in time generation, when block are shown in screen or carved, with just in time procedural query. therefore unnecessary unseen data isn't generated at all until it's pertinent.
Another trick would be path find from the camera position's chunk, to generate each chunk using a kind of frustrum filling djikstra by view connection to each successive chunk, so that chunk not seen aren't generated, given the cubic cost of generating voxel (memory and process), this overhead might help culling a lot of unseen data. I'm trying to determine an early skip for chunk if empty (maybe a two step generation, at chunk frequency then at voxel frequency), but I'm blanking.
I'm sure I hadn't think this through lol just typing down to keep note before forgetting it.
Sorry to use the comment as a scratch pad for sudden idea!
Well, feel free to try it. The lack of free rotation, scaling, and pattern breaking seems like a big drawback.
Which one are you talking about lol
If it's teh local voxel, they are are design for just that.
The rambling is about more general idea that pop up and have nothing to do with the idea of local voxel mesh.
But yeah I'm thinking about it, I need to start programming them, but I need to implement half edge and convex hull before that, and there are not related except they are part of the same project (spore ish open world creature phase).
My suggestion is to keep things simple and do it the standard way first, keep breakthroughs until you know you can do it at all.
You are very talented, thank you for this video, I learned a lot!
Man, this was so damn eye opening. Awesome video, thank you very much. :)
Hey, I have found this video really interesting. Thanks!
Subbed 5 min into the video. This is awesome stuff, thank you
Thank you so much, this helped so much. Also, very great analogy.
7:35 Gamer Dreams crushed
7:17*
This was super helpful in how I use height map and mesh in my game…
I just wanted to point out that the reason that Bethesda games don't use simple textures for roads and instead choose to utilize meshes doesn't necessarily have to do with game design. The Gamebryo engine simply isn't capable of texture rotation, and as a result paths and road textures can look jagged and strange when misaligned with the height map's vertices. You notice this a lot in Skyrim's many dirt paths. With a road's smooth, bending lines and edges, the only way to get them in-game is to kind of cheat and connect two road models which are cracked at the ends.
On the other hand, this does result in some very nice and rugged looking roads.
Nice video overall!
very nice explanation, i was looking exactly for this tysm
you have such a nice voice this was really fun to watch thank you
open world cities by indies...
clearly what we need procedural architecture
Modular packs with damaged piece alternative pieces so every building can be any size and have damage on any piece. I just downloaded a pack like this actually
Really interesting stuff. I didn't really know how huge signposting was, but now I'm really thinking about it.
It sounds like you're not a huge fan of proc-gen open worlds, but one of the takeaways I have here is that, if one is committed to procedurally generating an open world, then they should be generating paths and the core of their generator should be a lot more like a procedurally generated dungeon, just without the roof than any kind of statistical simulation of a world. At least, if the world is supposed to be as interesting as the AAA open worlds.
Every tool has its use. I'm not against generated worlds, but I found a lot of people making fairly basic mistakes with handmade worlds. Their strength is in things generated worlds haven't got figured out yet.
I agree with Perko, I'm big on PCG world but most of my focus has been about to solve basic design problem rather than automation problem, if you don't have solid design understanding, your algo will not do the work for you.
As a culture, pcg people are still mire into "magic math" perception of pcg with the occasional design "hack" like generating random critical path, but it's not there where they basically implement actual rules of design as generator. The curse of math pcg is that they only mimick the structure, they don't think about functional meaning of the structure or think about the structure of intent. So it's like having a sentence generator and all phrase are like "colorless green sleeps furiously", it's grammatically correct but mean nothing.
That's why you have so much focus on making thing look good and constrain it to be playable rather than algo that generate actual design goal to fulfill. In fact the way algo works attract more attention than what they actually do and why.
The poster child of this mentality is the article about tiny keep's random room, that just pile room on top of each other, use physics code to separate them, then construct a graph of room that connect to each other, then remove some line in cycle to have loop and dead end, that's the organisation equivalent of a spaghetti plate, why a room is there, why it has this size, how does the level flow and why, those are clearly not in the algo. It doesn't create meaning, it's only accidental. BUT THE PROCESS IS PRETTY TO LOOK AT!
Algo should be implementation of solid understanding of game level design. It also mean that you should have solid understanding, beyond simple intuition, of level design to begin with, something even in the game industry is a bit lacking.
Man! everytime I'm getting triggered by PCG discussion lol
Thanks for the clarification Craig, I appreciate that.
Neoshaman, I actually agree. That's the direction I think I was going talking about dungeon design instead of just dotting a world, but it looks like you're a few steps ahead of me. Yeah, you need grammar, you need a story, themeing, you need sightlines as part of your design, all stuff that hasn't been done yet in a really admirable fashion with algorithms. It will, of course, in time, but as you said, it requires an understanding that you can actually communicate in code.
But that all comes back to Points of Interest or SignPosts, and these are things that are fundamentally unique, where as PCG is fundamentally not. So, that's a thing. Unless, of course, you're doing literal signposts, because y'know, they're all the same.
"But that all comes back to Points of Interest or SignPosts, and these are things that are fundamentally unique, where as PCG is fundamentally not. "
I totally disagree about that, PCG has some fundamental limit but the reason it's popular is to make things unique. POI and signpost are like the easiest thing to generate.
The real problem is what I call te dictionary problem, which is multi faceted. At a simple level it's the bowl of owkmeal problem, all bowl of oakmeal are different, but still are essentially the same thing. The next step is the infinite chair problem, even though you can generate all chair in existence, it's still just a chair, it as limited function and expression. The next step is the expression problem, even though you have more than chair, you have to put them together in a way that makes sense and express something, you can have infinite room arrangement but only some have meaning, like the office rooms, the funeral room, the living room, etc...
The problem here is that even if your generator is expressive, this expression comes from human experiences, that's unique pattern to code, which comes back full circle to handcrafting things, because you need to encode specificity, human don't have necessary access to other human experience, I may know what a gallodrome looks like in martinique but not in america. So the process of filling a pcg algo is basically encode by hand those specificity.
Even though they are contain implicitly in the possibility space, and the algo can generate something we can recognize as part of the human experience, it's only accidental and would not be controlled. Since basically those element are description, they are like entry in a dictionary, hence the name. A large dictionary mean that you have good vocabulary to be expressive.
Expressivity would be the fitting range of an algorithm that match human experience, ie its ability to compose new description that make sense to human experience. So the irony is the opposite of what you described, you want pcg to be less "unique" in the absolute sense, as absolutly unique thing wouldn't be recognizable as part of the human experience. Hence why meaning and semantic must generated and manipulated to generate fitting composition. In another comment I talk about how I plan to create an algo that would make sense of sightseeing composition within pcg.
Hence why I compare pcg to a linguistic process rather than a mathematical one.
Also hierarchical generation is a quick and easy way to encode flexible meaning in a pcg algo. The highest composition capture meaning and intents as structure and propagate to lower structure. This mean that it's easier to generate using correlation rather than causation. Sightseeing and signposting would be notoriously hard if the terrain is generate and the right need to be found ... it's best to design the sight first and then generate the terrain to fit it.
The correlation problem can be easily illustrated with river generation, it's notoriously hard to find adequate river with enough control (for gameplay, sight seeing, etc ..) for a given terrain, but the terrain cause the river and it can lead to non predictable or unnatural composition if forced. We can reverse cause to correlation, since a river always flow down, it's way easier to fit a mountain to a river rather than the contrary, the slope and direction of the river is the same as the mountain, which allow better control as a river is easier to generate. Applying this to everything allow to generate complex phenomena, like npc interactions and history, without the simulation hassle we use to have. That's the basis of the infinite persistent NPC generation I'm planning to do.
That's pretty doggone fascinating. I deeply appreciate the linguistic angle, and I see what you mean about absolutely unique. I'm not sure how signposting could be the easiest thing in the whole process, since as I understood from this video, major signposts are essentially landmarks, and I'm a little iffy on a relatively trivial landmark generator.
As far as grammar and structure and rules, one thing that comes to mind is blockbuster movies. These follow a very solid and predictable three act structure. They also have a limited number of 'types' or genres. People continue to go back to them, however because of the large 'dictionary' and the potential to find rare 'uniques' which may not be truly unique, but are unique to their experience because of how rare they are, and yes, how common they are to the human experience. That kind of dictionary, color palette theming, juxtaposition, and a hundred other tools of sound, visuals and possibly even mechanics would be quite an achievement. We saw what it took for the No Man's Sky team just to create a visual grammar that was appealing. I'd love to work on the kind of algo you're talking about... but I have to admit, it would be a beast like no other.
Enjoyed that thanks. I was just thinking about the game Squad T the end when you said most maps are a bunch of trees although. I enjoy squad very much
Excellent video! I actually learned a lot from this.
7 minutes in & i've already learned more than my past 7 years
New England towns are actually laid out on ancient winding roads with nothing on the other side from the main street.
Thank you for the great video, genuine content.
Spyro or Crash. You can have an open world to explore but it will have less stuff in it and is mostly just flat or you can have smaller linear maps with maybe a little more up and down slopes but are filled with more objects and living things.
Compare Rayman 3 to Silent Hill 2 or Perfect World.
In Rayman 3 you can see from the introduction and first level that it's rich with windows and therefor homes built into the massive trees, garden gnomes and creatures littered around and small bugs flying around along with mushrooms and random holes with hills going up and down and bridges and hanging platforms with detailed background models and objects have small details added like lights.
Then in Silent Hill it's mostly flat empty streets with every little in them as well as each building being a separate level with the fog being used to cover the low capabilities. Perfect World is an mmo and like any mmo it has a lot of open areas with maybe just large trees and empty roads with specific areas highlighted with npcs and other objects
Not a game maker (not got a head for making code) but I found this very interesting.
Also helpful for when it comes to making mods for Bethesda games :)
OblivionSilk You actually barely need to know code to create a game. Of course it helps, but creating a game is often just a bit more complicated than creating a game :)
"creating a game is often just a bit more complicated than just creating a game." - Finduz11, 2017
This was a very information dense video that you've made flow quite nicely by using Bethesda's signpost pathing. Is there a good guideline of the variety of content that can be found at the end of the signpost rainbow? I'm applying a lot of this information to my exploration driven Dungeons & Dragons game and following the principle of breaking up patterns (like your point on the dike) it seems that I should probably consider other things to place at my signposts other than "empty, monster, special, trick, trap or treasure".
For a tabletop game?
If you know your players, figure out what things they might like that the other players might be a bit disinterested in, then add that as spice on top of a more fundamental reward. For example, finding a magic rune in a goblin's pocket, or a diagram, or a widget, or a friendly talking squirrel, or a strange pipe begging to be smoked.
By having the right player find them first, you can prevent the other players from putting the brake on.
Just be sure that you fold it back into the story with other finds: you want to keep the right number of loose ends. Too many and things follow apart.
In terms of physical signposts, it's mostly about the flavor you give when you describe it.
That's pretty solid advice, especially concerning loose ends with more story-focused campaigns. One of the things I find about location design is that it can be difficult to actually give them satisfying longevity in play. Most of the time you are "clearing" content and after the first couple of times you deal with signposts, they're only really useful as navigational markers.
That's easy: have them change hands, or the situation change radically (hurricane, plague, invaders, gold rush, new king, etc)
You deserve a like , very Informative .
I thought the thumbnail was a real world picture until I realized it was from fallout 4...
I don't quite understand how developers are able to work on the same asset of map simultaneously. What kind of procedures or pipelines are used so that you can have multiple people editing a single file at the same time without any of the modifications contradicting one another or corrupting?
Well, it depends on the engine, but the answer usually that the map isn't one map, it's quite a few separate maps that get stitched together.
One of a kind video Really. Good Work.
"Open world maps for open world games" PERIOD
idk that made me SHOOK
We need more of these
This is super interesting and informative, thank you!
Unifying the meshes and height map would reduce the vertices, but it would make texturing a nightmare, right?
It can't reasonably be done. The meshes are variable-density and the heightmap is uniform-density.
Awesome, I feel like I learned a ton of new stuff!
What's with the jittering clouds in Fallout at 5:50?
I'm pushing Fallout way, way past where players can take it by moving too fast and flying too high. The skybox is struggling.
Nice video! Could you make video about Battlefield map ? I'm just curious how they limit their map but it's look like huge map.
I don't have a console, so it's hard to talk about the Battlefield games, I think.
if this was a 10 part video . i watch it all. :)
great insight! :D cheers for the vid!
This video was oddly addicting
How to get around needing tons of building meshes for cities: MODULARITY
I'm using modular meshes with procedurally generated buildings for my game that takes place in an abandoned city. The buildings have randomized numbers of floors, facades, side-brick work, windows, chances for roof accessories, awnings, air conditioners, decals for damage/ graffiti... not hard to set up, but it can get a little tedious. I bit the bullet and bought a modular building pack with 5 types of buildings with mix and match ground floors, middle floors, and top floors, but most of the big world meshes are my own (roads/ sidewalks/ building greebles like pipes and vents).
Nope!
Modularity isn't as efficient as you think. I go into more detail about that on one of the NieR videos.
Nice work!
Do more videos like this! : D
Is this the Unreal Engine 4?
Nice vid. Can you also talk about their poly count for most of the object?
As time marches on, that concept doesn't matter as much. These days you can use LOD meshes and it's all much cheaper than it used to be.
@@CraigPerko Thx for the reply, but I have to disagree. When you are making basically 4k openworld, you would want to be really careful about the poly batches you are loading at one time. Maybe one day when we have super computer it won't, but that is decades to come.
This video help so fricking much!!!! Thanks!
what software do you use to explore games like this
He used the noclip console command in fallout
Any advice when your game character can fly and cover a lot of ground quickly?
That's a complicated situation because not only are things moving rapidly, but you probably can't block their sight. I'd focus a lot more on having long, medium, and short-range POI.
world look like developers got crazy and scaled everything in random way, that is why it's so random.
thanks alot, very nice video
Thanks Dude. Interesting video
I would love to see video about terrain streaming in Unity and about performance and all the problems it creates for developer.
Unfortunately, I'm not an expert on that.
Seems like I have an extraordinary sense of direction then hehe
“An open-world game but there’s no game”
Breath of the Wild in a nutshell.
What is name of this software applications. Please tell me
Well, it's mostly a video game called Fallout 4. And a little bit of Unity.
So how what is a good size hieght map to use? What is a good FOV? How tall should your buildings be? How do you deal with big floats?
All of these things depend on your project. If you're just getting started, use a small map: the entirety of Fallout 4 is small enough that you'll never run into floating point errors.
Smallish height maps are fine - 1024x1024 or 2048x2048 should work fine, because that can be a kilometer across without any problem. If you need larger or denser spaces, you should use multiple height maps and only load the ones you need or your performance will suffer.
As for buildings: they need to be tall enough to do their job. Whether that job is acting as a soft barrier to let you split one side of the building from the other, or maybe that job is acting as a tower to climb, or as an instance gateway to a dungeon, or as a standalone point of interest to sack... design the building to do the job you need it to do.
Thanx but what I don't understand is how many pixels I stretch the 1024x1024 map. how big should it be in world space floating points? It cannot really be 1024x1024 because my screen is bigger than that so what size is it really? for example if I am using a triangle strip with the 1024x1024 height map - how big should each triangle be to represent a kilometer?
You don't have to build the terrain mesh yourself, but typically a pixel is either one meter or half a meter.
I recommend using Unity's built-in terrain, or forking out for something like Gaia if you can afford it.
I am building the mesh myself but I did some googling and figured out that the "scale factor" is what I was asking about and that its up to the programmer to pick a scale factor for their world.
Yup! But I recommend 1 meter or 0.5 meters per pixel.
Is that unity or not cause I need help building
It is possible to make the game world based on the vast sphere?
It would be pretty hard.
I was watching this years ago and than Unreal 5 and they use mesh (Nanite) for everything. :)
In world of warcraft a lot of the mountains are pure height map.
As well as in Morrowind, iirc... Now jump forward 10+ years...
That's probably to save some space in your hard drive.
How can they have that much geometry in the world at once? Is it done with an LOD-system?
They have an aggressive LOD system that uses occlusion. I think it also does predictive loading, which is why my high-speed, high-altitude antics caused loading pauses. Unity can pull the same basic tricks, they're not unusual.
I see, thanks. I've played around with LOD in Unity a while ago but I couldn't find a way to make it unnoticeable enough. Maybe I should look into it again..
You absolutely need LOD to do things like this. I generally use a simple decimated version of the mesh, it's usually pretty hard to notice if you decimate it the right amount.
I gave it another go and it came out better this time. I think the quality of the meshes is crucial... I tried the mixamo decimator which seems to work quite good for organic stuff. Also the Blender decimate modifier can exclude a set of verts which is nice :) Now I'm thinking about ways to integrate LOD level generation in my modeling process for the more hard surface models. Do you have any workflow tips you would share?
The easiest way to do it in Unity/Blender is to simply use linked duplicates and add a carefully-tweaked decimate modifier to each. Named correctly, Unity will automatically put them in an LOD object, and because they're linked, modifications will propagate.
Hard surfaces do well with planar decimation.
From any site I can download these homes bro
Brilliant video
I'm a newbie at this can you point me at a place to study this for beginners
Not sure if there is one.
"If you're following the icons, this might not be where you would go." ...and that is what has ruined every Bethesda game I've played since Oblivion.
How much time it took to build this big map and city
Fallout 4? A team of dozens of people took several years.
@@CraigPerko thank you
And a question can we make (single person) can make GTA 5 like game in a year
@@AREA-wc3qm Nope! But if you scale back and use a lot of shop assets, you could make something smaller and simpler.
@@CraigPerko ok thanks you
@@AREA-wc3qm it took thousands of very talented people a few years to make gta v.
craig you are awesome i love the video
good video thanks for sharing
If you want to see indie game open world city, try hobo tough life
did you make all this by yourself?
No. Bethesda did. The flip is that you can. It just takes a lot of time. especially making buildings with working doors texturing, baking etc. Very few people are taking worldcreation as seriously as the man in this video but they deffinately should. You've gotta be very consious while doing these things. I do so myself and have found that the difference is night and day. It also takes a lot of time to really understand the programs you use but it is worth it if you''re a creative person.
Interesting!! Thanks.
my wife was killed...long ago...my son was kidnapped...long ago...when waked up the second time I was so angry, I jumped off the cliff...searching for a sniper and forgot my family
Good talk
I was taking a look at GTA V world using CodeWalker, it pratically doesn't have terrains, at least heightmap terrains... It's all meshes, you can turn on the wireframe and you will see that entire topology of the models comes together in a very harmonic way, roads, terrains and buildings, like they were modelled together, it's mind blowing! They probably has some very different workflows and ways of doing things that unfortunately they never disclose... :(
Very nice video by the way, it helps a lot and it brightens a lot of questions I had regarding how fallout 4 world was designed. Thanks!
GTAV uses more advanced generative techniques. Look into procedural modeling for that kind of thing.
Will gaya work for the Unreal Engine?
I don't know, but I'm 100% sure UE has similar capabilities.
No - Gaia is a Unity asset. I'm sure that UE 4 has similar capabilities, like Craig said. UE 4 has many beautiful environments
Great Video!
What about open world maps for driving? Like GTA.
Line of sight is even more critical.
24:23
*Pulls out a gun*
Oh, interesting...
By the way nice vid! I clicked this vid to check out the map, and I noticed it was just a clip to learn about making games. Before I pressed the back button, I started to get interested, so I sat back and started watching ;)
Can't say I disagree since I'm only one person, but my experience of signposting and blocking in fallout 4 is very different to what you portray and what Bethesda intended, at least after Concord, the feel of almost being railroaded right to Diamond City, the feel of needing to shun any landmarks because they're so limiting to ones movement (especially in combat), the feel of almost always being told what a site is from miles off. As said I don't think I'm in the majority nor even intended player base, but for me, it just makes me stick to the wilderness they want me to avoid unless I'm being railroaded to the next quest.
Fallout 4 had some serious weaknesses, and the biggest one for me was the HUD icons that encouraged you to not explore. That said, I didn't feel any of the other things you seem to have felt - exactly the opposite. I didn't even get to Diamond City until ~15 hours into the game.
Some setpieces were very rigidly laid out, but the world as a whole didn't feel that way to me. I'd be interested to know more about how you felt, because it's clearly different enough that it should be taken into account.