3D Models and Terragen - Basics

Started by j meyer, July 03, 2015, 03:17:21 pm

Previous topic - Next topic


That was why i mentioned the ngons before Michael.They can be problematic in nearly in all programs.


You mean in addition to quads right, not that it can't happen with quads? Anyway at least for me, Ngons are much worse problem in TG. entire areas will not render at all with ngons, so whole sections of models missing. At least that is what happened when I first had that problem in TG2.
It has been eaten.


July 27, 2015, 11:43:59 am #62 Last Edit: July 27, 2015, 12:38:08 pm by Kadri

You can have problems even with quads or trigons using in different programs and with normals etc.

If i remember correctly the only place you will have no problem with ngons are  flat surfaces,not sure.
And it depends on the kind of surface angle,normal etc. So it works sometimes without a problem even in Terragen.

But ngons are so bad it is even meaningless to search for a solution for this if you ask me.
The safest way is whatever you model try to don't use ngons.

Others might have more knowledge about this.

j meyer

Ngons are a problem,most definitely.Even on flat surfaces.

Non planar faces can be problematic indeed in conjunction with 'smooth normals'.
That's what I was refering to in Michaels thread in my reply to Matt's post lately.
They - non planar faces - are hard to avoid when sculpting.

Going to do that shader related stuff tonight,maybe I'll add an illustration for the
non planar faces as well.


Well if at some point you do break all this down into a wiki entry of some kind. there should be a section/paragraph on things not to do/things that never work. So Ngons and how TG just can't cope with the trauma, and some of the other things that are fresh in the threads right now. Just saying. ;D
It has been eaten.


July 27, 2015, 09:29:24 pm #65 Last Edit: July 27, 2015, 09:33:19 pm by Kadri

J meyer this for example worked without a problem :


With flat surfaces i did mean this kind of examples.
But when 1 or more vertices changes position (in this example in the Y axis for example) for whatever reason, it can get problematic very fast.
I haven't tested it with different normal angles.So even it works like above it could be problematic with different normal settings as you said.
I would subdivide it before serious use and get rid of any ngons.


Here's a short informative link: http://blog.digitaltutors.com/ngons-triangles-bad/  I didn't realize ngons and triangles should be avoided (too new to modeling  ;)). When hauling through Poseray, it's automatically triangulated, I think.


July 28, 2015, 11:20:39 am #67 Last Edit: July 28, 2015, 11:22:41 am by paq

Well the blog is talking about subdivision surface modeling. If you prepare assets for Terragen, you don't really have to care.

When working with subdivision modeling, you actually prepare a 'cage' model, that will be subdivide and smooth out using (most of the time) a 'catmull clark subdivision' method.
And this subdivision function gives a better visual result when dealing with quads, or at least predominant quad models.

But not every models have to be made that way, if you model a little architectural piece for you environment, you can just mix poly, tries, ngons ... it doesn't matter.
Behind the scene, there is only triangles anyway ... quads and ngons are just 'helper' for modeling, because it's easier to deal with.

For terragen, you only have to triangulate ngons before export. You can leave quad as quad, or you can triangulate everything.
Just save a 'working' model too in you modeling package (using native format, because .obj sucks :P), with the original topology for further editing.


j meyer

Kadri I can't see what you have done there without seeing at least a wireframe.
Maybe there are exceptions.

Modeling in quads is the way to go,although an occasional triangle won't hurt.
Ngons are to avoid.
As Matt pointed out earlier elsewhere on these forums TG triangulates any
imported model anyway and you as a modeler should only triangulate your
(finished) model if you prefer one triangulation way over the other.
Which can be necssary on some occasions.I'll show an example for that later
when posting the non planar stuff.Hopefully by tomorrow.


Quote from: j meyer on July 28, 2015, 01:06:52 pm

Modeling in quads is the way to go,although an occasional triangle won't hurt.
Ngons are to avoid.

Again, it depends of the context. There is no need to model only with quad just for the sake of it.

'Only quad' is needed for a 'subdivision modeling workflow', where the mesh will be deformed in animation (skinning, morph etc).
Actually quad is not the only thing you have to deal with, you also have to care about vertex valence, but that's an other story.

If the model will not be deformed (often called hard surface modeling), then you can use n-gons too. As long as it's looks fine in the viewport, who cares.
In the past some software didn't accept to subdivide ngons topology, but it's not the case anymore.

If you are not using subdivision surface on the top of you model, then there is no rules. Of course n-gons should only be
used for flat surface , and maybe you need to add some edges here and there to help the internal triangulation to resolve
concave geometry, but that's all.


July 28, 2015, 01:55:38 pm #70 Last Edit: July 28, 2015, 01:57:12 pm by Kadri

It depends like Paq said where the ngons are used.
On a flat surface they get triangled without problem mostly.
Terragen triangled the obj itself here for example without problem.
But with displacements and normals etc. used in Terragen it can get ugly.
So in Terragen with displacement i would avoid them all the time.

Here is the file itself with the object.

j meyer

July 28, 2015, 04:16:43 pm #71 Last Edit: July 28, 2015, 04:26:33 pm by j meyer
paq - This is about modeling/models and Terragen!
        I at least try to keep my info related to that.

Kadri - The last time I tried that was with that moviemodels train wagon,if you remember
          and that did not render the large flat surfaces.
          There might be exceptions or Matt has changed something in the meantime,don' know.
          Will try your file and tell here.

j meyer


I would like to see the old model you mentioned. Curious too.
Probably the vertices weren't at the same axis-surface or displacement was used if it doesn't worked then.
But maybe Matt might have made changes too of course.