3DS Max Help: Modelling for games Thread

Hi everyone
I have been away from the scene for a long time after finishing a games course at uni, and i am now hoping to get deeper in to game development (lone wolf style). I will be learning how to use 3DS Max, UDK and CryEngine which will take a long time to get the hang of again.
Unfortunately i have forgotten some important things and needed some basic info before i begin. Can some1 please answer these questions?

1. When making a model for a game, do all the polygons have to be joined/connected at all edges to remain stable in the game engine?
2. Can polygons/objects overlap or go through in 3ds max and be stable in a game engine?
3. Do rogue vertices cause any problems after importing a model to game engine?
4. Do all polygons have to have 4 edges or can you have a mix of 3 and 4

thanks for taking time to read this. will appreciate any help, and once i learn everything i intend to learn, i will help as many people as i can : )

Edited by: Assassim

1. No polygons don't need to be connected at all edges. There are some exceptions though, for example a collision mesh.
2. Yeah they can. You should avoid doubles though, they can cause render artifacts and other problems.
3. Stray vertices are a result of unclean modeling. Any software has ways to check that, should do that as a standard routine before export!
4. All exporters for game engines I know do triangulate the polygons anyway, so yes 3 and 4 edges. Just n-gons don't go (more than 4 edges).

BTW, meshes don't disassemble just because they're used in a game engine!

Edited by: BogOfEternalStench

Thanks for ur reply Bog, really helpful.
So would i be able to simply place one object in the middle of another without using a boolean?
Also i heard that game engines dont like booleans, is that a load of bs?

Edited by: Assassim

You can place objects in overlapping fashion, can cause collision issues though sometimes, especially when model is very complex or multiple parts overlap eachother. May depend on the individual game engine as well. If you can achieve it by continuos modeling, i'd always prefer that. But for a tree for example it doesn't really matter wether the branches are connected to the trunk or just plugged in.

Bolean operations do often cause very unclean results, especially with higher polycount models. They need a lot of tweaking. You have to avoid doubles, non-manifolds (edges from which more than 2 faces do emerge), strays and so on. I personally wouldn't use them for making game assets, try the modeling way, poly by poly.

Is there a certain model you have in mind?

Edited by: BogOfEternalStench

Thanks for the info mate
As for the models i am going to create, i am going to create many such as guns, buildings, maybe a vehicle or 2. bits of foliage and characters (high res and low res). I am using Eat3D tutorials to learn how to do the advanced stuff, would u recommend them?
Its going to be a project to work on in my spare time so i can get fluent at these programs again. i haven't got a concrete idea yet though, but i will probs be posting regularly on here from now : )

I see. Yeah, Eat3D, Digital Tutors, Gnomon is definately quality stuff, can't go wrong with that. I'd register at an advanced forum as well, such as polycount or 3dtotal. Polycount is focused on modeling for games, you can learn a lot over there.

I want to add something:

1. There is no technical reason, you could have your polygons floating however you want. If you plan to animate the model joining all the vertexes/edges that should be joined is a must. Collision models have stricter requirements although I'm not an expert on which engine requires which.

2. That is generally considered bad modeling but sometimes it has it's place like modeling complete mouth with tongue or just saving polys on low-poly models. Just make sure you get the rigging and textures right. Usually if the two polygons are on the same exact plane and overlap most engines produce a tearing result where both of them flicker when the engine does the sorting process each frame.

3. Is there reason for them, like an insertion point for props? I have personally avoided using them.

4. You can mix them since it's all going to get triangulated before the game BUT. As we know good edge flow is essential to good modeling and that is why you need to always always keep your model in quads and keep quad working version as well after you export for the game. Going back to quads from tris is never easy although I'm not sure if 3DS max has some feature to that. Still, you should use tris where you make shapes that require it and for low-poly modeling you should also make the seriously non-planar quads into tris so you can control which way the 2 tris are formed to ensure good result both in shape and texturing.

about the triangles, it's all about where you place them and at what point you are in your workflow. If you want to subdivide your mesh a triangles might cause you trouble so It might be better to keep quads, but later on reagrding the in-game mesh it's not that important.

As a rigger I hate to see triangle in some place but love to see them in others
Polycount.com

Here's a good article about smart triangle:
Wiki.polycount.com
If you look at this picture you'll notice triangle in the shoulder are as well as in the elbow...

Yes. Basically you should use the triangles to support logical edge flow of quads while eliminating subdivisions where you don't need them

Also note the palm and thumb. Making the thumb without triangles is pretty pointless.

Thanks for the info guys. I didn't know much about the triangle business, the article about topology and using triangles in the right places is really helpful, ill be looking deeply in to it when the time is right. BTW i was just curious about the stray vertices.

I started modelling an aircraft using a tutorial The poly count is ridiculous (almost 2mil) What would need to be done to make this an acceptable game asset, i imagine its a long precess?

Also it is now time to put the materials on. i have objects that are turbo smoothed and i used swift loop to create some hard edges. now i want to take off the turbosmooth so i can use Unwrap UVW easier. If i delete the turbosmooth modifier, apply the unwrap uvw, and then apply turbosmooth again, will it remember where i used swift loop?

Edited by: Assassim

Assassim wrote:
I started modelling an aircraft using a tutorial The poly count is ridiculous (almost 2mil) What would need to be done to make this an acceptable game asset, i imagine its a long precess?

Wow, that tutorial must not have been made for a game asset creation :p Well depending on how you achieved such a high polycount and what game platform you are targeting it's not necessarily a long process. You should get rid of unnecessary subdivision and there could be a lot of them based on the polycount. Or it could be just details that you want to preserve but in that case you need to make a lower polycount mesh and bake those details onto a normal map. Can you give us a wireframe screenshot of the model so we could take a look at it?

Also it is now time to put the materials on.

I would advice against doing much texture unwrapping before you have the mesh itself closer to what it should be. You can plan how you're doing it, think about seams and find reference images for texture baking etc., just don't work on unwrapping yet because your polygons will likely change around quite a bit.

Edited by: shadowflar3

Lol nah it was just a high poly modelling tut by digital tutors. id prefer game modelling though because i want to get in to the games.

How would you make a lower poly count mesh out of it, do you just keep deleting the uneccessary sub Ds and then apply normal maps?

I think the model is complete now, i have gotten to the materials part on the tut.. i saved a seperate version with no smoothing as well
im having troubles posting pics up i have them saved on my comp. i will edit this shortly and post them

Assassim wrote:
How would you make a lower poly count mesh out of it, do you just keep deleting the uneccessary sub Ds and then apply normal maps?

Depends on what you got and where you want it to be, sometimes you can automate the process a bit. If you are way off in polygon count (like need to reduce polys to <1% of what you have) the smartest thing is to model a new lowpoly mesh on top of your hi-poly model by hand using the other mesh as a reference. Afterwards you could use the hi-poly mesh to bake a normal map for your lowpoly mesh to add detail.

But it all depends and I'd have to see the mesh first

Edited by: shadowflar3

High Poly Version (around 1.8m faces)

Low Poly Version (Around 5000 Faces)

I saved the model separately before i applied any smoothing, will i be able to use that as the low poly version??
I did follow a tutorial to make this so i cant really say it my own model completely. The tut was Digital Tutors - Introduction to 3ds Max 2014 and it
doesn't mention anything about modelling for games, texture baking, low poly models etc

I'd say the polygon amount of 5000 is about right for some PC FPS game where you would have only couple of them on your screen in addition to the player and environment. For RTS where you might generally have like 20 of them as an unit it would still be a bit much. So you must think of what kind of game and engine you are designing for and how will the model be used. Do I need to model the cockpit interior? How many sides do I need to have in this cylindrical shape so it doesn't look blocky from the distance it is viewed from? Do I need to have to model air intake holes or gun barrels or could I get away with just using a hole texture? Will the player ever see the model from this side?

Game asset creation is somewhat different from modeling for renders where you can have subdivisions just-in-case. You have to constantly keep polycount in mind so things like turbosmooth for roundness and swiftloop for sharp edges are often just temporary tools after which you need to adjust the result towards the final model.

I recommend you start small and import your models in a game engine to see them in action. Besides bug testing I find it keeps me motivated and gives me ideas on what to work on next.

So if I bake some maps including a normal map. Could I get that low poly version looking a bit more like the high poly or would I need to model a new low poly model? Would a normal map make parts like the blocky cockpit look more smooth/rounded also?

Would a normal map make parts like the blocky cockpit look more smooth/rounded also

They can affect the appearance of the surface, but they don't do anything as far as the silhouette is concerned. So you'd still be able to count the number of sides on those wheel things, for example.

Is your facecount (5000) reported in quads, or triangles? It can be the difference between 5000 and 10000. For reference a lot of games seem to be falling into the 10-40k triangle range for hero characters (with the exception of RTS, mobile, etc type games). So unless you have a smaller target in mind you could spare some extra polygons to smoothen out areas like that.

Assassim wrote: So if I bake some maps including a normal map. Could I get that low poly version looking a bit more like the high poly or would I need to model a new low poly model?

Not really if the difference between the 2 of them is just the subdivision modifier. Normal map creates details like creases on clothes, bumps and indentations but it won't affect the mesh geometry itself like subdividing surfaces does.

Would a normal map make parts like the blocky cockpit look more smooth/rounded also?

Yeah I was wondering about that cockpit. It seems to have enough polygons for smooth appearance but you can still see the edges of polygons. Are you sure you have set those polygons as smooth shaded instead of flat and checked that the edges are not marked sharp? Hard to tell because jpeg compression creates a bit of a mess there. You might want to find out about smoothing groups (if you haven't already) when you start to adjust smooth/flat polygons and edges to make your model smoothing look the way you want. Check the link at the bottom.

Anyway it is common to find that some parts do need further subdivisions while others are fine. That's when you need to make the modifier only affect those parts of the model to prevent unneeded subdivision. I'm not sure how it goes in 3ds max but in Blender for example I separate the model into 2 meshes, apply the subdivision modifier in the other mesh and merge the meshes back together.

Some reading on normal map baking: Wiki.polycount.com
A small article/tut on smoothing groups, for Blender but mainly because the first image says it all: Blenderartists.org

Cryrid wrote:

Is your facecount (5000) reported in quads, or triangles?
For reference a lot of games seem to be falling into the 10-40k triangle range for hero characters

Its reported as quads so yea it would be roughly 10,000 triangles. I was thinking this vehicle would be for an fps game, or a 3rd person indie game where the player will control the ship from an outside view like unreal or halo vehicles, but i still need to do a lot of work on the the low poly, haven't had much time due to work hours. I think I need to reduce the polygons in certain areas and increase them in others to make it more acceptable as a game model.

Do u think if i kept it at around 10k triangles, it will be suitable for a modern game like that?
So am i right in thinking that in-game models for modern action or adventure games have only smoothing groups, and nothing like turbosmooth for roundness? What was meant by adjusting the result of swiftloop towards the final in game model? I had a nightmare of manually deleting 1.8 million edges after reading this

shadowflar3 wrote:
Are you sure you have set those polygons as smooth shaded instead of flat and checked that the edges are not marked sharp? Hard to tell because jpeg compression creates a bit of a mess there. You might want to find out about smoothing groups (if you haven't already) when you start to adjust smooth/flat polygons and edges to make your model smoothing look the way you want. Check the link at the bottom.

The viewport was set to realistic but when i set it to shaded, the cockpit appears about the same. I will read up on the links you sent, and will check up on smoothing groups. I have used them a little in the past but i always had sh*tty results. Maybe i wasn't doing it right.

I am using this model to gain more knowledge on the whole game modelling subject. I will probably create a more simple object or character to put in udk engine after i learn everything, either that or I'll adjust this one for it
Thanks for helping me out with this ppl
A drink for y'all

Edited by: Assassim

Its reported as quads so yea it would be roughly 10,000 triangles. I was thinking this vehicle would be for an fps game, or a 3rd person indie game where the player will control the ship from an outside view like unreal or halo vehicles, but i still need to do a lot of work on the the low poly, haven't had much time due to work hours.

Well, 10k polys is not that much for modern machines if you are positive you only have 1 of them on screen at a time. But still I suspect there are some ineffective polygons there that are creating the high number without adding detail. It kind of still is a thing that needs to be fixed as models need to be efficient so the computation power is freed for shaders and particle fx etc.

So am i right in thinking that in-game models for modern action or adventure games have only smoothing groups, and nothing like turbosmooth for roundness?

Nope. Game models are just fixed polygons so you will have to apply all your modifiers before exporting the model.

Some of the key things that will help you are smooth shading (adjusting vertex normals) and smoothing groups. I found you a 3DSMax tutorial which will hopefully clear any confusion on smoothing Try it on the cockpit.

Youtube.com

Edited by: shadowflar3

Do u think if i kept it at around 10k triangles, it will be suitable for a modern game like that?

Easily.

The aim is that if you're using x triangles, then the model should appear to be in that general neighbourhood of detail (bonus points if the model ends up looking like it should have a lot more polygons than it actually does). Otherwise 40k+ isn't a huge deal in modern fps games as long as those verts are being spent well (80-300k for a vehicles isn't even unheard of, in vehicle-oriented games). There's always LOD models so that vehicles in the distance can use lighter meshes.

So am i right in thinking that in-game models for modern action or adventure games have only smoothing groups, and nothing like turbosmooth for roundness?

I don't think its something that is terribly common, but games like Call of Duty Ghost are starting to show more use of realtime subdivision (Youtube.com)