Hi all,
Newly registered to the forum, so a proper noob in every sense of the word.
The thing that got me to register are the challenges set on the forum based on concept art. As an aspiring 3d artist I have been dabbling into the fine arts and teaching myself, slowly but surely, how to model basic geometry and how to learn to texture.
The one thing I keep running into, and perhaps I have not researched this correctly, is that the recommended approach is to maintain quads where possible and to try and work out tri's within models.
Now in saying this, the advice that I have read has no correlation towards the gaming industry and how this translates there. When looking at a lot of the models produced in the monthly challenges I see people using tri's where they can be avoided (flat surface) and fail to connect vertices together (floating quads in geometry).
Now really, my questions are:
Should quad be used over tri's ?
Why do most game engines report on tri's rather than looking at quads?
Is it OK to have a mixture of both?
How would this translate into game engines (having more than (Example) 4 tri's connecting up to the same vertices?
I know my questions are very basic, but then so is my knowledge.
I hope someone is willing to dish this noob some serious case of the knowledge!
Either way, thanks all, and keep on inspiring!
Replies
1. you "should" try to stick to quads purely because they subdivide cleaner, they're easier to read for other artists or animators, and UVing can be easier with quads. however, there's really no harm in using triangles for your final low poly mesh because:
2. engines report in triangles because they're mathematically the smallest face they can report on, you can't have a face with two sides. this makes it quicker for renderers to compute. the reason it's good to at least look at the triangulation of your final lowpoly before exporting (using an edit mesh modifier in max for a non destructive example) is so that you have final control over the direction of triangulation... auto triangulation can lead to janky normals.
3. yes
4. read above.
Ultimately, all polygons can be broken down into triangles, using quads helps keep the distribution as clean as possible. you can get weird normals at verts that have an number of outgoing edges other than 4 (quads) or 8(triangles) these are known as poles, you shoudl avoid these if and wherever possible.
I have been looking at some models, which seem to have flat surfaces, however I see people opting to still slice the quad into 2 tri's. I know its probably a loaded question, but could you perhaps tell me why someone would opt to slice up quads into tri's ? Does this not require additional calculation?
The reason I am asking, is because I want to go through all historical 'Community Noob Challenges' and start creating based on those concepts.
When I look for example at the August one, I see something similar like this
Props to 'JustMeSR' for creating the low poly, however if you look at some of the flat surfaces it seems there are unnecessary splits in the quads.
Could you tell me when someone should make such a consideration?
Thanks again for your prompt response, proving the polycount community is a great one even for those who do not posses the skillset
In the end (for an engine) it will be always tris so it doesn't require additional calculation. You can't prevent it being calculated as tris. Sometimes you need to do this to have a better control of how the faces are split not leaving it to the engine. Normally you wouldn't triangulate until you are finished if not needed as it is easier to work with quads, but once you are done you can triangulate the mesh and check how it looks (as almighty_gir already explained) as it might affect the shading
This at least would be the reason for the mesh you are talking about.
Example time!
The finished mesh
Triangulated within Maya
Unity's automatic triangulation
You can see only a few instances where they triangulate the mesh differently and in this case it doesn't matter too much, but on more complex meshes that will employ normal maps it could lead to shading errors.
The take away is that you don't /have/ to triangulate your mesh before bring it into an engine, but it could cause you problems if you don't.