Can we use Triangles in Game Props ??

Chidambhar_Swaroop

Hello Lovely Polycounter's

While modelling a low poly for a high poly model, can we have more tri's instead of Quads? I am speaking in a game development perspective not animation or visual effects... Mainly on props not chars or envi. Once you import anything to a game engine, it automatically converts it into tris, so what is the use of creating it in quads ? 

I got this question because I read this article from 80 lvl - https://80.lv/articles/the-difference-between-a-concept-and-a-production-mesh/ . If you look at his production screen shot, first of all, it has a lot of quads, it could have fewer quads. In the modelling package itself, I can do it in tri's and reduce the tri's count whr it is possible, right ? Which will help in performance.. I think he was speaking in a movie perspective but I am just asking how it is supposed to be for games. 

For eg look at this both picture A and B

Picture A

Picture B


In Picture A, I modelled it in quads and it has total 94 tri's but in Picture B, it has only 45 tris... Is it ok to do like Picture B? Should I keep something in mind while modelling ? for eg I can't connect so many tris with a single vertex or something like that ??

Thank you for the help

HAWK12HT

@Chidambhar_Swaroop you are right, game engines are fine with tris so long that particular part is not animated etc. The article you shared deals only for films and animation industry. 
Game industry: tris, ngons all good to go as long as hi poly bakes fins, non deforming body.
Film/ TV / Ad industry: 99% quads is a general rule of thumb

Bek

Once you import anything to a game engine, it automatically converts it into tris, so what is the use of creating it in quads ?

Try working only using triangles. In particular, try selecting edge loops on a triangulated mesh. There's your answer.

Regarding pictures a/b, this is less to do with triangles vs quads as it is general mesh optimisation. Generally speaking, an 'optimised' mesh will not have any unnecessary geometry that does not contribute to the silhouette or lowpoly shading. Whether or not the mesh is made of quads or triangles is irrelevant because as you touch on, everything is triangles to the GPU in the end. So B is more optimised (but not perfect) because it has less geometry, not because of the ratio of triangles to quads. You typically work with quads because it's easier, but there's nothing wrong with using n-gons or triangles. In some other contexts (Film; particularly meshes that deform) they do sometimes require quads only (Or so I'm told) but in games that's not true at all. Characters might have some specific topology requirements around bits that deform like elbows and knees to stop things twisting and stretching in a horrible way.

When you delve a little deeper into optimisation you'll learn more about how vertex splits and good triangulation (along with other considerations like UV usage) are more important than your final triangle count. You can read a thread about vertex splits here.

Larry

Triangles may produce artifacts in animation mostly, and some smoothing/lighting problems, since a triangulated mesh may not have straight lines. Also, you cannot automatically multi-select edge loops or rings if u work with a triangulated mesh, giving you extra time to select every individual edge of it. As people above me mentioned, triangulation is only a means of optimization technique to convert 2 edge loops into 1 and remove unecessary polygons, mostly used in geometric shapes.

Chidambhar_Swaroop

Got it guys... Thanks for the information .....

poopipe

You need a picture C 

It is usually more efficient to try and avoid long thin triangles and poles where many tris meet at the same vert - this is to do with the number of tris you need to draw per pixel at various viewing angles/distances 

  In this case you could get a better result by simply turning edges on B to make the tris more regular. 

kanga

Jusst to answer the thread title. Yes. Depends on how you are going to use the model. If you arnt going to use subdivision on the model and you are exporting it to an app like zbrush, where you will do a high rez dynamesh for polishing, you can have any configuration you like. If you are making a model for baking its a good idea not to allow any long tris, or strangely turned ones.