Actually I am an artist who mainly focus on 3D graphics and animation. Recently, our team members decided to make a mobile 3D battle sci-fi game( similar with Infinity Blade) with some programmers who are just graduated. In our game, we got 2 characters and 1 background. They said that each character should be around 600 poly mesh, and the total poly count should be under 1500 poly mesh in the same scene. I really want to know is that right or not? As I am an animator, I know that 600 polygons for one character is really ugly even I apply a normal map on that model.
So, please let me know the answer and thank you so much for your times!
If made a game and the characters work for around 1000 polys on one mesh. It depends on your mobile platform, what you use etc. There are lot of tutorials and articles about that issue, just google it.
Some of the most gorgeous models I have ever seen have been sub 1000 tris. Not polygons, tris.
To note though, 1500 for the whole scene sounds rather low. I’d say to just work everything low/mid poly and not worry about things too terribly much. Just keep testing each iteration to make sure it runs smooth.
Also, if you have Pro, you get access to occlusion culling, which will help performance (any objects not visible are not rendered, and thus potentially allowing slightly higher poly counts). Keep in mind as well, frustum culling is built-in to both versions, and that keeps anything that’s not within the camera bounds from rendering as well.
So, a lot depends on which mobile devices you actually want the game to run on. The page at:
http://docs.unity3d.com/Documentation/Manual/iphone-Hardware.html
shows comparisons for various iOS devices. If you are a new team, consider this:
a) Decide how much GPU time you want to spend on each character. Perhaps that’s 20% of the frame time.
b) That leaves 60%, so perhaps 50% for the environment
c) then you have 10% for effects and GUI elements.
I’m not saying those are universal constants that all games use. Just giving you a way to break things down into chunks. So, if I am not wrong, your two animating characters can be rendered 5 times on screen and make your frame rate. If each character takes 20%, you can render it 5 times in 100%. Also if the environment is 50% you should be able to render it split screen through two cameras and make your frame rate.
Knowing all this, you can get your low spec device (maybe a 3GS) and work out how many triangles you can have. Note, that the GPU performance will be massively dependent on the shaders you use and how many draw calls you make.
I meant to add that it’s infinitely easier to take a game and make it more shiny shiny on high end devices, rather than make a game you’ve built for a high end device run on a low end device.
I have a +600k poly scene running smooth on a iPod Touch (5th gen) with around 200 draw calls… if you target newer devices poly count dosen’t really matters much anymore. I’d say, if you use a mobile/unlit shader you can achiev the same performance with ten times the polycount as when using a diffuse shader.