Hi, I'd like to procedurally generate models of some things I like, both larger things like diverse terrain, landscapes and geology that can look like anything in the real world and maybe even fantasy, but also smaller things like vehicles, architecture and mechanical stuff. I love seeing the different shapes, sizes and colours of these things and I want to become good enough that I can create something beautiful with them.However, at this moment I know almost nothing about modelling and what programs and methods to choose, they each have their technical details and applications that I can't parse. I fear that I will miss something that would work better for what I want or choose the wrong thing and only finding that it won't work after dumping a lot of effort and time in it. My only experience is playing around with some toys that model just by manipulating basic shapes with codeblocks. This was very inefficient so I want to stop fucking around and get serious. Eventually, I'd like to make a game or something like that with these models, so I will also have extra constraints and I'll have to choose something that can work well with Unity or maybe another engine.The problem, and the reason I come to you, is that I am mentally retarded so things like UV mapping and defining triangles frighten me. They also seem incredibly time-consuming when combined with procedural generation and I'd really like to be able to make something that is at least decent as fast as possible. I will learn the harder stuff eventually and I'll start with it if it's necessary, but right now I'd really like something more "high level" and easier to use. Could you please give me an overview of what methods exist for this kind of thing so that I can understand what options I have and what would work best for me?
>>1031585You should take my advice because i am also a retarded amateur. The short answer is it depends on what kind of game you're making and what you need the assets to do. The long answer is-There are a million youtube procedural terrain tutorials, and blender's geometry nodes arent THAT hard, so you can just mash perlin noise together until you're happy and call it a day if you really want. Perks of that is it'll probably tile easily if you want an open world, but it does limit what you can generate a bit. You can always add to it manually thoughSculpting onto a grid is a good idea. Takes a bit to get a feel for it but once you do you can use tools like locking the Z axis or dynamic retopology to make nice shapes that are actually quite efficient. Its the approach i like personally at the moment, but it can be a bit annoying some times. You can probably also do some terrain generation in-engine, or predefine modular terrain tiles. Once you have that you can use geonodes and other shit to cover a surface with random permutations of trees and foliage, but thats about as far as ive gotten. Uv unwraps are actually fine, as is extruding triangles. You mostly just let the computer handle it>this map? Smart project itAnd then you just shade it however you want. Speaking of, shaders are engine specific so you'll have to remake them in unity. Idk if it has a visual shader editor or not but fuck shader code
>>1031588>it depends on what kind of game you're making and what you need the assets to doBasically it needs to be generated from some kind of code/prototype that I create and can change based on certain parameters. I also think it needs to keep some similarities to normal models, so that things I can learn from others that work on normal models also work on mine, though I understand that that would raise the difficulty.Basically, something like a vehicle prototype that generates multiple different models for different terrains, technological advancements and things like that. I guess I like thinking of things that way, shapes that follow certain functions and constraints rather than a single thing. Also, if it works it would also give me a lot of different models and I really like the variety and the fact that I can tie it to other parts of a scene or even some gameplay. >you can just mash perlin noise together until you're happy and call it a day if you really want.>Once you have that you can use geonodes and other shit to cover a surface with random permutations of trees and foliageYeah, that's probably how I will start, but I think I will need something that looks better and is more easily controlled so that I can make crazier stuff with it, kind of like pic related. It would also probably be difficult to use that for the smaller stuff, like vehicles and the like.>There are a million youtube procedural terrain tutorialsThat's why I want to see all the techniques that people here know about, so that I can understand this whole thing. Also, maybe it's just because I'm a beginner, but what little I've seen of terrain generating tools, and 3D pieces of art in general, that look very good so I really have a lot of competition here.
>>1031598for the time it would take to make a procedural vehicle generator you could basically just make them all by hand. And how many do you really need anyway? all up GTA vice city only had 100 vehicles of any type, including bikes, planes, boats, helicopters and RC toys. you can probably do it, but what's it really going to be doing? starting with a car template and stretching parts of it? you can just stretch it yourself with far more precision and flexibility>i want to make a cave citythe cave and the city arent that hard to make seperately, but you'll struggle to make a cave city generator that can generate any other kind of city, and im not sure how many cave cities you'd actually need in the first placeit might be ironic to say on /3/, but i really dont give a fuck about game graphics. define your project first, then build the tools you need to complete the project. for all i know you're making cookie clicker, you might not need any of this and if you dont then you can save a lot of time and effort
My idea is that, if I understand and define rules that dictate the shapes of certain things thoroughly enough, I could create a larger variety of models than with normal modelling. If I also manage to make those rules loose enough, I might even make something that is a bit surprising and interesting to myself, I would really love that.There are other advantages too, it would feel less limiting and contrived than having a limited number of normal models that can be customised only up until a certain point. It would also be easier for models to be interacted with and changed.In a way I think that, if I understand what I'm modelling well enough, I should be able to integrate it in my generator without some absolutely insane, out of this world time investment. For the variety of models that I would get, it might just be more efficient than normal modelling.There's also the fact that terrain generators seem to already have some of the qualities I'm looking for, so this might not be impossible, though I understand that terrain is far easier to do than models.Of course, I don't know much about 3D modelling so I realise that I might be wrong and completely delusional on all of this. I understand that I might end up just making normal models with some customisation, but I think I should at least give it my best shot. Also I'm not making a game now, I'd like to first become able to at least make some decent 3D art.
>>1031621>>1031623I'm dumb and didn't link your post.
>>1031623Its possible, but you cant do it. You dont understand how to describe the problem, so its impossible for you to actually solve it.>ill just describe the rules for how the shapes of cars are madeWhy dont you think about what that actually means for a minute or two?
>>1031650>Why dont you think about what that actually means for a minute or two?I know how that sounds and maybe I am delusional, I know I won't reach the Platonic ideal of any object, but I think I could still make something nice. It's a problem of how well I can figure out the logic for what I want to create, not a technical issue with implementing my logic into some program. Researching the problem more led me to believe that I should be able to do something like that out of simple shapes and boolean operations. It seems a bit hard, but reasonable and I feel like I should get used to it relatively fast. If I can't make it work well enough, then that is my problem and not something I'm asking others for help with. The problem I have now, is that working with either SDFs and then converting them to meshes, or working directly with meshes, results in issues with the meshes that can't seem to be solved automatically. One way might be automatic retopology if it's actually good enough, and most people seem to say it really isn't. Either way, I should be able to use just SDFs for things other than game assets which is nice.
>>1031655It's not clear to me what your problems are, but maybe you should just start and work on getting just one thing done to a level you'd be happy with. We don't know if you want a photorealistic car, a stylized car, a low poly car or just a blob with the label 'car' floating above it. You haven't told us what your expectations are, what your level of expertise already is and what exact problems you're facing. Without that, talking about it is quite pointless and you should just do the thing and when you hit a wall try to solve it yourself and then come back, ideally with screenshots of what your problems are, or what progress you want to show. From what I've read from you, you don't really seem to have any idea what you're talking about though, so go and prove me wrong
>>1031659>what your level of expertise already isI've said that I know basically nothing about modelling. I've only done some polygon boolean operations in some browser toy-programs a few times.>From what I've read from you, you don't really seem to have any idea what you're talking about thoughI really don't, that's kind of why I tried to ask a broader question. I thought that seeing an overview of what methods there are, and roughly what is possible with them, would give me some basic info and let me choose a place to start at.>but maybe you should just start and work on getting just one thing done to a level you'd be happy with.Yeah, that's probably a good idea.>We don't know if you want a photorealistic car, a stylized car, a low poly car or just a blob with the label 'car' floating above it. You haven't told us what your expectations areThat's true, in a way I'm too much of a beginner to know exactly what to say. I don't know if this makes sense or not, but the geometry and texture detail on the trains in the OP picture, from that same distance, would be nice as a start.>It's not clear to me what your problems are, I wanted to see a kind of overview of methods of procedurally generating 3D models to get an idea of what would be right for me.Also, the methods I found out about, procedurally generating meshes with polygon boolean operations, seem to have some massive issues. I wanted to know if I can avoid those issues or bypass them entirely by using some different technique that I do not know about.
>>1031664The method depends on what you want to generate. So pick something (grass, road, railroad, landscape, building, city, tree) and search a tutorial to get a feel for it
>>1031655>It's a problem of how well I can figure out the logic for what I want to createYes. What logicical rules exactly are you expecting to use to define the shapes of vehicles?
>>1031664You don't really need to boolean, you could generate each simple shape (cylinders, bent rectangles, spheres, cones) and just have them be, no need to boolean them.SDF has the benefit of having basically infinite quality, but it's much harder to find tools, and you'd probably have to write some SDFs yourself, which you probably won't be able to >he geometry and texture detail on the trains in the OP picture, from that same distance, would be nice as a start.Uhhh... Good luck... 'as a start'...As a start you could put these models into a AAAA game and make billions from sales, you know? As a start.
>>1031665Yeah, I was overwhelmed by all the videos that people can put out about this, but that's what I'm going to do.>>1031698>Uhhh... Good luck... 'as a start'...Yeah, you're right, that's too much detail just for a start, but I feel like at least the general shape, the bigger elements and a simple linkage system on the wheels should be doable.>no need to boolean themWhat would happen if I needed to make a hole in a certain shape?>>1031694I don't know exactly how to express myself, rules that, based on certain parameters, like length, height or vehicle type, would generate a model that has those attributes with some variation in it's shape.
>>1031706>certain parameters, like length, height or vehicle typethese arent even enough to change the shape of the vehicle. you cant describe the rules, so you cant possibly write a program using it.
>>1031712There would be way more parameters that would imply cutting, combining, transforming shapes and other things like this. The program itself isn't the problem though, if I can't write it then that's that, I'll try something simpler.
>>1031706>What would happen if I needed to make a hole in a certain shape?you wouldn't. You could create the shape with a hole in the first place, you could put together multiple shapes in a way that forms a hole.But as I said before: just start, instead of talking about it vaguely and half baked. I assume you'll quickly start having problems to talk about instead of grand ideas
>>1031717cut where nigga? i dont give a fuck if you can or cannot cut, you dont know WHY a car is car-shaped, so how are you gonna tell a computer where to make the cuts?
