Triangle count for game characters

Hi guys.

I’m modeling a character for a game and I haven’t really been able to get an idea about how many triangles the in-game model should have. I want it to be as optimized as possible yet still have enough detail to look good. Does anyone here know how many triangles the character models in Skyrim have for example?

At this point my character’s triangle count is 165 984 and that’s way too high right? So I’m going to create a low-res model and then bake the detail into it with a normal map. But I need some kind of target as to how many triangles I should aim for for the low res model that’s going to be rendered in the game.

Thanks guys.

Give an example of what you think “low res” looks like.

Modern games use anything between 30k (30 000) and 10k (10 000) for high to mid poly models.

I think the Skyrim player model and some NPC models sit at around 10k and weapon models at around 2-3k.Try to use textures to fill in the details rather then modeling them in.

By “low res” I just mean the model that ends up in the game which will have a much lower triangle count than the sculpted, fully detailed model that I have right now. I would like it to have about thesame detail as Skyrim. I will bake most of the detail (for example wrinkles in the clothing etc.) into the game model in a normal map so they won’t be modeled. But I chose to model certain things like gun holster straps around the legs. I might remove those and use textures instead. Not sure yet.

Reason I’m asking about Skyrim is because I’m interested in the fact that they have the base “naked” character model and then they have the hair, armor and weapons on top of that. Do they somehow make the parts of the naked character model under the armor and helmet disappear when the armor is on or are those parts still being rendered?

If any of the existing models are available in a form Blender can import, download a few of them, import them and check the face count. Also take a look at how they’ve organized the mesh. Some games use characters that have separate legs, arms, shoulders, heads, etc.

Right now your model does sound like it has too many faces. But! if you go for a mesh that is sufficient to display your details through normal maps and textures, you may run into texture stretching problems when the mesh is animated if the count is too low. It’s a balancing act. Find some characters that are already working, and use those as models for your own effort.

Well I reduced the subsurface modifier from 2 to 1 and now the tri count is at 47 840 which is much better. I’ve identified the hands as the biggest offenders. The difference between the subsurface modifier off and on is 20 000 tris! I’m going to have to clean the hands up and somehow keep the shape right. Also, I can remove the subsurface modifier altogether from some parts like knee pad straps etc. and it will still look good enough for in-game.

On the subject of how the mesh is organized. My character right now is in parts. The head, helmet, goggles, hands, body, shoes, vest etc. are all separate objects. What is the simplest, most hassle free way to organize it as far as rigging, materials and textures etc. are concerned? Should I keep the parts separate which would make it much easier to texture them and keep the materials separate. Or should I combine them all into one object which would probably make animation easier? I’m worried that if I combine the character I will have to assign different materials to specific faces to make it look right, but will that still work once I import the model into Unity? Will Unity know which materials are assigned to which faces?