Questions from someone coming from 3D for advertising...
polycounter lvl 6
Hi! A friend in the gaming industry told me about Polycount recently and told me its a great place to learn and ask questions regarding 3D for games. I've dabbled very lightly in 3D for about 8 years but only started taking it seriously about a year or two ago. I use Maya and have been learning as much as I can. Im currently interning as a 3D artist at an amazing digital media company here in NYC focused mainly on TV/internet advertising. Ultimately I would like to get into the gaming industry and have (very) recently begun learning the differences between 3D for games and 3D for everything else. With that said, I have some questions which I haven't found direct answers, most likely a bit novice, but bare with me! 
- Tris or quads or both? I see a lot of meshes that are all tris...some quads...and sometimes both?? What's the verdict on this? Varies studio to studio?
- Booleans. I know people hate them...I tend to use them when I can and clean up the geometry. What's the general consensus in the gaming industry? Is it more 'model however you want as long as it gets done in the end' or is there protocol? How do I know is something is too high of a polycount?
- I see a lot of pieces people post here of props. What's the general lighting rig used? Standard 3 key/point/fill? I tend to use 3 area lights and a backdrop for all my portfolio pieces.
- I know mental ray is off limits (correct?) so do people render just using the standard Maya/Max renderer?
- Material nodes. I've always used mia_materials in Maya...but since theres no mental ray in gaming, do people just use the standard blinn/phong/lambert etc.?
- Texturing and UVs. Lets say I have a model of a robot I've built...is it good practice using just one texture sheet and lay EVERYthing on that texture sheet? I've been doing it piece by piece and that's probably not correct as I have to match the resolution for each piece with one another.
- Low poly modeling. I know a main thing is to model a high poly version of something, bake out the diffuse/normals etc. and throw those on a low poly model...but if I'm building something with hard surface modeling, should I just bevel the edges to avoid sharpness so I don't need to sub-divide in the end?
Thanks for your time!
- Tris or quads or both? I see a lot of meshes that are all tris...some quads...and sometimes both?? What's the verdict on this? Varies studio to studio?
- Booleans. I know people hate them...I tend to use them when I can and clean up the geometry. What's the general consensus in the gaming industry? Is it more 'model however you want as long as it gets done in the end' or is there protocol? How do I know is something is too high of a polycount?
- I see a lot of pieces people post here of props. What's the general lighting rig used? Standard 3 key/point/fill? I tend to use 3 area lights and a backdrop for all my portfolio pieces.
- I know mental ray is off limits (correct?) so do people render just using the standard Maya/Max renderer?
- Material nodes. I've always used mia_materials in Maya...but since theres no mental ray in gaming, do people just use the standard blinn/phong/lambert etc.?
- Texturing and UVs. Lets say I have a model of a robot I've built...is it good practice using just one texture sheet and lay EVERYthing on that texture sheet? I've been doing it piece by piece and that's probably not correct as I have to match the resolution for each piece with one another.
- Low poly modeling. I know a main thing is to model a high poly version of something, bake out the diffuse/normals etc. and throw those on a low poly model...but if I'm building something with hard surface modeling, should I just bevel the edges to avoid sharpness so I don't need to sub-divide in the end?
Thanks for your time!
Replies
afaik, people generally avoid them because of the crap topology you generally end up with. But if you can clean it up, and find it faster that way, then go for it.
There's... about the same standard as there is with ordinary photography. Sure, there are preset lighting setups that photographers may use depending on their subject matter or what effect they're trying to get, but ultimately they're using lights to model their subject matter; to emphasise the forms they want to show off, and to dull the less flattering bits. The presets may take you much of the way, but knowing _why_ you're using one over another is the key to beautiful lighting. I believe game art is no different.
The point is to use a real-time renderer, just like ones that game engines use. First prize is to use an actual game engine (UDK, Marmoset, CryEngine, Unity, etc.). Otherwise, it's a viewport capture using some sweet real-time shaders. Offline renderers (like mental ray, or the Max/Maya scanline ones) are avoided because you can achieve effects, lighting, shaders in them that aren't possible in a game engine in real time.
What shaders you can use depends a lot on what game engine you're using, what your target hardware is, and whether you've got the support of graphics programmers and tech artists for shaders you're looking for. If you're comfortable with constructing shaders (like using Maya's hypergraph), you could look at using the UDK's material editor to build some of the shaders you want. Blinn/phong/lambert are most common, but you can achieve much more than that if you're willing to build your own.
It depends. If you're going to have a bunch of objects (e.g. all the pieces of one robot) appear together very often on screen, then it's a great idea to put them all on the same texture. If it's a robot that'll have parts swapped out, and there are many, many parts, then it may be better to do each swappable part on a separate texture. (Your programming team might want to do some fancy batching where it dynamically builds the texture atlas based on what parts are needed, say.) In general, the fewer texture fetches the better, but as with all game art optimization, you're balancing improvements in one part of the pipeline with sacrifices in another.
In general, if you can't see the difference from your game camera at some kind of "standard" distance from the object, if it doesn't make a difference to your silhouette, if it doesn't help with deformation, and if it doesn't help with your UVs, then get rid of it. Whether you'd bevel something or let the normal map do the work (or both) is kind of dependent on that, I feel.
You can subdivide your highpoly. You shouldn't be subdividing your low though... I'm not sure I understand what you're getting at.