Basic questions about video game assets pipeline

Worraps

Hello, 

This is my first topic here, I hope I am not doing wrong by posting in this section for my questions. 

So I am currently learning 3D, as I want to be environment artist in videogames. Despite the incredible amount of online ressources, I still have somes specific questions I would need answers to move on and continue my learning. They are various and about different subjects, and for somes I found different answers, so that's why I came here, to be able to have some specifics answers to theses specifics questions. By advance thank you for your time reading this and maybe answering these following questions : 

1) First, at the modelling step, is it okay to split my meshes instead of trying to extract the shapes of it ? Let me explain, if I want to model a table ( I model in blender ), is it better to add first my top part, then in edit mod add each legs as a new cube and put it at the right place, or is it better to add loop cut on the top part then extrude the legs from it ? This is a really... basic question but I could not find a clear answer anywhere as some peoples add details to their model by adding new mesh in edit mod and some others try to extrude them. 

2) Still in modelling : I've understood that quad modeling is the big rule in 3D, but in videogames every model are triangulated. I've read it was due to mathematics and how the game engine handle calculate easier with triangles than with quad. But when do I triangulate my model ? Do I need to do it with blender ? At the very end of my worflow just before importing it in the real-time engine ? Also, what about N-gons : i've seen many hardsurface worflows based on booleans, but as you know once you apply them you get terrible topology. Does it really matters if every thing gets eventually triangulated and not even animated ? ( I am talking about environment props ). As long as your shading is ok, are N-gon ok ? I am talking about what is accepted in industry since I will use your answers in my portfolio. 

3) Is there a "acceptable" threshold of polycount for environment pieces ? Since you model in quad, this number is doubled when you triangulate it. So is there somewhere a sort of table saying what max number of poly you can use for let's say, first, middle and third order of props importance in the screen ? Same for texture resolution : 2k, 4k ?  and I know all of this is also related to LODs, but when you present you portfolio as a junior Environment Artist, do you need to show LODs of your model ? 

4) Speaking of portfolio : when you are self-learning, what do you base your models on ? On artstation, I see many artists creating based on concepts arts, wich is I guess how it works for some big companies when you work as 3D artists. But what if you don't have concept art ? Do you imagine one ? I have a loooot on artbooks on many games, films etc but I guess I can't just take a concept art that I like, make a game-asset based on it and place it in my portfolio without asking to the guy who made the concept ?  I know you can also model things based on real life objects but this must also be related to the company you are applying, and what kind of games they are making...

5) Last question for today : is general, is it better to have one texture set for all your model ( even if that comes with some parts having less definition) or you can have two or three texture set ? Maybe that is more adapted to animation and in videogames you have to stay with one ? 

So that's it, I know that is a lot but even if you have the answer for just one of these, I'll really appreciate it ! 

Thank you for reading

Worraps

Eric Chadwick

1: Usually ok to split models. Construct them in parts, as they would be in real life. 

An exception is for normal mapping... these models can look better in-game when modeled contiguously. http://wiki.polycount.com/wiki/Normal_Map_Modeling#Low-Poly_Mesh

2: triangulate before baking a normal map and exporting to your game engine. http://wiki.polycount.com/wiki/Texture_Baking#Triangulation

3: Look at examples on Artstation from artists working on games. Also check the sample assets from Epic's Paragon. Demonstrating LODs is unnecessary... For entry level it's much more about artistic quality than it is about technical issues like LODs, which are often created automatically anyhow.

4: using concepts is accepted practice. It's polite to credit the concept artist, asking permission is not required but is often appreciated.

5: depends on the size/importance of the asset. Again, check samples from Epic and elsewhere.

Good luck, and keep the questions coming!

Ghogiel

Worraps said:

1) First, at the modelling step, is it okay to split my meshes instead of trying to extract the shapes of it ? Let me explain, if I want to model a table ( I model in blender ), is it better to add first my top part, then in edit mod add each legs as a new cube and put it at the right place, or is it better to add loop cut on the top part then extrude the legs from it ? This is a really... basic question but I could not find a clear answer anywhere as some peoples add details to their model by adding new mesh in edit mod and some others try to extrude them. 

That can be fairly normal to tackle models as several parts. There can be pros and cons, situations where you do and don't. The way to reason why do it one way or another might be what the model is supposed to do, ie is it a highpoly subD model, lowpoly, How low poly, does it need to animate, ease of solving the modeling, is having separate parts going to be annoying in the UV or baking process, time etc 

Worraps said:

2) Still in modelling : I've understood that quad modeling is the big rule in 3D, but in videogames every model are triangulated. I've read it was due to mathematics and how the game engine handle calculate easier with triangles than with quad. But when do I triangulate my model ? Do I need to do it with blender ? At the very end of my worflow just before importing it in the real-time engine ? Also, what about N-gons : i've seen many hardsurface worflows based on booleans, but as you know once you apply them you get terrible topology. Does it really matters if every thing gets eventually triangulated and not even animated ? ( I am talking about environment props ). As long as your shading is ok, are N-gon ok ? I am talking about what is accepted in industry since I will use your answers in my portfolio. 

Everything is a triangle. Even ngons are triangles. You kinda answered the big part of the question yourself, does the shading work or not?

That's not the entrie argument, as there might be other concerns dpending on the function of the model, and obviously you don't go out of your way to make and model in explict triangles or ngons, as many modelling tools work best with quads and so on, but it's all tris at the end of the day even if you don't connect the edges.  But tri and ngons happen, and they might work or not in that situation.

As far as triangulating, there are few sticking points where there this is a problem in any workflow where the model has to move from one app to another though> Different apps triangulate the model in different ways, so unless you force collapse all polys to tris some problems might occur. Most apps are basically dumb to the fact that everything is a tri and remove any control from the user for no reason. So we end up with a problem of apps using triangulated models under the hood, but they make up and remake that tri topology on their own in different ways depending on the app.  

As a general rule the time to sort it out is when you are going to bake the model or exporting it to bake. If there are no baked normal maps involved it probably won't be of as an issue, but you could just add a triangulate modifier or whatever the app does just before export to the engine as a best practice, so you are sure nothing happens to triangulation by some other process along the way. FBX exporters will have options about triangulating, which you could use too.

These different workflows with ngons, booleans etc, all do specific things and are for different purposes. Some are undoubtedly for high poly modeling that has nothing to do with the actual models you might send to a game engine, but are only used to bake maps from. Maybe it's just a rock and the topology is a decimated mess of tri and the low poly was never even quads anyway. There are tons of workflows doing a load of different things for different reasons to different ends, which may not be a general "this is how you model for games" type of workflow. Maybe it's a "this is how you quickly make a complex surface that shades decently, but has shitty topology and is only good for baking maps, and sucks for unwrapping or just about anything else" workflow. Nothing wrong with that either. Different horses for difference courses.

Worraps

Thank you for your detailed answers ! Honnestly you both gave me the kind of answers I needed. But to make sure I understood : a viable worflow as environnement artist ( I say environnment but let's go with props, and modulars kits to build scenes ) would be :
1) Modelling or sculpting the high poly mesh, no matter the way ( relatively speaking ) as long as it looks good and without shading issue. Spliting in different logics parts is accepted but can create some problems when baking normals, so it has to be made wisely. 
2) Then I have to create the lowpoly mesh, this time with the best possible topology
3) UV Unwrap
4) Triangulate it before the baking of the HighPoly on it to avoid vertex normal flip and shading issue later.  
5) Baking
6) Texturing 
7) Ready to go 

I know you just told me there are differents worflows for different situation, but I need to have a kind of a "guide" to follow along, and to know what I have left to do, a kind of checklist. I've been learning Blender with different courses, experimenting a lot, but game dev is a whole new world ! So now that I am learning Subtance Painter, I'd like to follow along the course with a model I made and not the one that come with it, in order to eventually put it in a real time engine and ( I hope ) get my very first game-ready asset. That's why it's so important to me to know very precisly what steps I have to go through ! 

Again, thank you for your answers. 

Worraps  

Eric Chadwick

Here's a checklist if you like. http://wiki.polycount.com/wiki/Texture_Baking

Worraps

Hello,

thank you for the link, very interesting things and exactly what I was looking for. There are some step here I never heard about before so really nice checklist. 

Worraps

icegodofhungary

I think people have answered your questions directly, but I would like to offer some advice that are indirectly related to your questions.

Try not to start out by making a whole environment. It's pretty impossible to learn several pieces of software at once, while learning several skills, while learning your art fundamentals, while trying to build a portfolio. You're just not going to be able to do it all in one go. Start simple, start basic. Build your foundations one brick at a time and you will be much better off later on when you do tackle the big stuff.

Start with modeling single objects. Pick something in your house, something you can put your hands on and view it from different angles. Teach yourself how to make real objects look real. Don't even worry about the texture part at first. Just focus on form (a fundamental of art!) and proportion. Learn to make the software do what you want it to do. Learn your modeling tools. Then start learning how to UV and texture.

As you get better, you'll learn what you can model as separate pieces, and what you can't. You'll learn when ngons and triangles get you in trouble. You'll also learn what you can get away with in terms of topology, not everything has to have beautiful wireframes.

What I wouldn't do in your shoes, is try to impress Naughty Dog, Insomniac, Blizzard, Sony Santa Monica, etc with your first object or scene. You're setting yourself up for frustration by putting that much weight on your shoulders. Focus on the fundamentals, develop strong skills, and then worry about getting a job. Your portfolio 2 years from now will not have your first object in it, I guarantee it. Nor should it. As you get better, you'll remove older stuff and put your best work forward. Don't worry about picking your projects based on your employer just yet.