Retopology for game assets question.
polycounter lvl 10
I have been building high res sub-patch models and then manually retopologizing them with 3D Coat for reducing poly count and baking normal maps. Since poly counts aren't as critical these days I'm considering just using the subpatch model with a division of 1 for the retopomesh. This creates a much cleaner and accurate mesh but with extra polys. Would this be an acceptable retopo mesh?
The low poly mesh has 761 tris.
Replies
after it is in your game or render if you saw that it is taken up such small screen size that multiple triangles occupy the same pixel then you could reduce it or create lods.
There are ways you can measure what performance impact a thing has and making those measurements is faster than redoing your work. So, it makes sense to test and see if there is actually a problem first.
Then you can weigh the decision like this: how much of a difference does manual retopology make in terms of performance in the final product versus how much additional time does it necessitate?
In other words, if your scope of responsibility is "make a good looking prop", just find the path of least resistance to get that done. If you have some notion that doing things the hard slow way is always better, what if you are wrong? Then you've charged somebody a lot of time for no benefit.
If your scope of responsibility is also to know what performance impact will eb of the art, then you have to learn to use the profiler tools in whatever engine the thing is rendered in. And then it's just a simple test - run test without the thing present, run test with it present. See what numbers changed.
Thank you so much for your comments Alex_J! I'm trying to create high quality and realistic props for a portfolio so I wouldn't be able to determine the needs of the game or use case of the prop. My goal is to produce assets in a reasonable amount of time without cutting corners. Thank you for the profiling suggestion, that is something I can spend time using for these.
Cutting corners would be to make an assumption and not do test.
Just because you spend extra time doing things a manual way does not mean the result will always be better. Some times yes, sometimes no. Only way to know is to test.
So if you do take the time to run some test and compare difference between this model and one that has been manually retopo'd, probably its not a bad idea to mention results you found in the breakdowns (if you are showing breakdowns.) But I don't think that's really necessary to do for a portfolio. Who knows, I'm not the expert on that stuff.