If you don't mind, I wanted to pull a post by you out of that dumpster fire of a thread down there and repost it here:

Whilst MD uses delauney triangles, using a modeling app, in my case blender, it is a lot easier to use quads. I tend to model in very large quads and make use of mirror and subd modifiers, the large quads allow easier UV segmentation and it is much easier to make sizing morphs if you only have to select one or 2 large quads and move along the z-normal, with the Subd this makes the morphs much smoother than trying to do it with the subdivided mesh. I even use this method if I have to apply the modifiers as I keep an unapplied copy that I use as the binding object in a mesh deform modifier.

...I agree, perfectly. I always wondered what big advantage anyone saw in doing tris (or even majority-tri meshes), unless they work in the game industry where that's pretty much required (game engines even list their polycount budget (scene, world or old-school POV, a'la Unreal) in tris by default.)

Everyone else is welcome to salvage and repost/expand on modeling techniques, especially WRT Poser/DS. IF you still stick to tris (or, well, whatever-the-NURBS-converter-spat-out), I'd be interested to hear why...

Don't some modeling apps allow you to convert quads to tris? I'd have to look it up, but I believe Blender does that.

EClark1894 posted at 10:06AM Wed, 05 February 2020 - #4379110

Don't some modeling apps allow you to convert quads to tris? I'd have to look it up, but I believe Blender does that.

Yes you can do that in blender very easily, what we are discussing (thanks for the rescue @Penguinisto) is whether you should.

The process DURING modelling is much harder in tris. Things like wanting your program to select a line around an arm won't work, in general most tools in your modelling program won't make sense or simply won't work.

Last I've looked into game 3d, they convert into tris well into the mesh creation, they don't start like that. Things could have changed since I looked into that, though.

Dyn-cloth engines may get better results with a triangulated mesh. May. A big piece of fabric felt down to ground for example. But tri-mesh != tri-mesh. One has to find out which routine gives the best result in a certain case.

That does not mean that you need to construct your model as a tri-mesh from the start. You may make a tri-meshed copy from your quad-mesh just before going to the dyn-engine (I think that's what Earl meant).

Afrodite-Ohki posted at 6:47AM Wed, 05 February 2020 - #4379115

Last I've looked into game 3d, they convert into tris well into the mesh creation, they don't start like that. Things could have changed since I looked into that, though.

No, I should have clarified; the result has to be in tris, not the process, and this is for faster/easier GPU consumption. In the bad old days of the 1990s, where, say, 200 polys was an absolute max budget per mesh in games like Quake (err, you could cheat it, but...), you basically constructed the whole thing in tris whether you liked it or not, if only to squeeze out as much detail as you could.

Penguinisto posted at 11:04AM Wed, 05 February 2020 - #4379138

Afrodite-Ohki posted at 6:47AM Wed, 05 February 2020 - #4379115

Last I've looked into game 3d, they convert into tris well into the mesh creation, they don't start like that. Things could have changed since I looked into that, though.

No, I should have clarified; the result has to be in tris, not the process, and this is for faster/easier GPU consumption. In the bad old days of the 1990s, where, say, 200 polys was an absolute max budget per mesh in games like Quake (err, you could cheat it, but...), you basically constructed the whole thing in tris whether you liked it or not, if only to squeeze out as much detail as you could.

Ah, the lovely times of Lara Croft's pyramid boobs.

One fat bit of trivia to toss in like a monkey-wrench... GPUs prefer tris.

(Still prefer the thing in quads, though... makes 3rd-party morphing easier to happen (even though most folks are lazy in DS these days and use collision, custom morphs are still quite useful). Also, I suspect it makes the texture/normal/bump/spec maps easier to work with for folks who do add-on textures (not 100% sure, but I suspect it would be..)

(I'm skating at the limits of my knowledge here, so corrections welcome if I'm in error!) AFAIK the advantage of triangles is that they are always flat - the three vertices will create a plane, so the polygon is called planar. A polygon with 4 or more vertices can be non-planar if one of the vertices is not on the same plane as the others. Render engines triangulate geometry when they render so all quads and ngons are split into triangles, but a non-planar polygon can potentially be split in different directions. The render engine won't split that poly the same way each time it renders, so that can result in unpredictable shading artefacts and flickering in animations.

As already said though, the modelling process is much easier with quad meshes. Selections are easier, edge flow for deformations is simpler to create and manage, subD is more predictable etc etc.

I can make one specific use case for triangulated meshes in Poser though. As Superfly doesn't have micropoly displacement I have made static props using decimation in Zbrush which results in pretty scrappy triangulated meshes, but I aim for a polygon count that keeps a lot of silhouette detail. I fix any major errors, make sure it works with a level of subD, and use the texture maps to carry the fine detail. Using more geometry does take more memory when rendering, but then so does using 16 or 32 bit texture maps for micropoly displacement, so it's a trade-off that I think works for some uses.

Base mesh, Zbrush screenshot, 50k triangles -

Final render, Superfly -

You're perfectly correct in that a non-planar poly can split in goofball ways, often differently in the same render if you shift the camera just a smidge (which is often a neat trick to fix the stray firefly.) In such cases, it's hard to say whether it would be more advantageous to pre-digest something into tris or not before rendering it in Poser, but holy damn that would make for one real sluggish pipeline (compose figure and scene, export as mesh, triangulate it, etc.) DS neatly ducks this quandry by allowing you to triangulate a specific mesh right there on the spot (good luck fixing a bad split on the spot though, but TBH that's uber-rare and you can sort of fix it by various means). And yes, it's there for pretty much every reason you've mentioned and more.

I don't know if Poser has anything similar built-in to pre-digest a mesh into tris, but I would hope that it does, yes?

In Blender you can either select all the faces and triangulate them, or use the non destructive Triangulate modifier before you export if you do want a fully triangulated mesh for use in Poser or another application

BTW @caisson that is a nice render of the log!

Cloth simulator and tri-meshes:

https://animation.rwth-aachen.de/media/papers/2013-VRIPHYS-MultiGridCloth.pdf

"Wrinkle Meshes"

http://www.matthiasmueller.info/publications/wrinkleMeshes.pdf