As I was starting doing some testing I got surprised over and over about how sub-d behaved in version 2.63.0 (last version I tried was either 2.59 or 2.60 I believe).
Inside the the grease pencil stroke you can see what I’ve been forced to do. That edge loop that didn’t help to make sharp edges was instead used as an anchor for that extra quad and the pole it give rise to.
Also a quick test with creasing instead reveals that unlike before, creasing the boundary edge loop didn’t do anything, until I also creased a diagonal edge attached to a corner vertex.
Is this how I’m supposed to make sharp edges with sub-d from now on ?
It’s strange, i reproduced the model you are showing , to obtain sharp edges at the boundaries, i need to select not only the boundaries edges, but the 4 little diagonals ones there
Thanks for the experiment and reply. So this is a surprise for you also ? I’m guessing you haven’t (either) used latest version(s) for sub-d modeling for a while ?
No, in fact it’s just that i don’t do subidivision modelling directly starting with a plane, usually i first build the base shape of the entire model, and once i’m done then i add a subdivision surface and work from there.
So i would not have started with a plane on which there’s a subsurf, the shape would always end in having some kind of depth that require more logical (to me) selection of edges to be creased
By example giving a depth to your model , the edges i need to crease are then more logically these :
As you see in the crease color, i would not have needed to use those little diagonals ones that does not make much sense to me.
@MCollett
I think my memory played yet another trick on me, hehe!
**** happens when you’re away from 3d for almost a year, especially when you had only a few months of practice before it.
@Sanctuary
To test it out I extended (extruded) the model according to your image, with the top ad bottom edge loops, and the vertical corner edges, creased with value of 1.0. It did better but still not right. A relatively large area near the vertical corner edges appeared as if the crease values were a bit lower than 1.0. Even if I knew that creasing the edge loop of the octagon would’n help, I tried it with result as presumed.
I double checked all the crease values, and they were still 1.0. I tried removal of possible vertex duplicates, and checked the normals; I saw no flaws there. Couldn’t figure out the cause.
@All
In any case, is creasing something portable, that is, do other software support it ?
And lastly; do you prefer using crease weights or extra edge (loops) for the purpose ?
No, creases are not very portable. Edge loops give finer control, and would generally be preferred for a true high-poly model, but creases can help if you are trying to keep the poly count down.
It’s also problematic to work with creasing + smooth shading.
If Sanctuary’s model is switched to smooth shading (to even out the shading of the “bowl”), the shading of the crease-defined corners will be very messed up. The shape gets tightened up, but the shading is still averaged out between the “floor” and the “wall”.
Ah yes i forgot to mention it sorry, i always use an Edge Split modifiers wen i use Subsurf + Crease, it’s an old habit that i have done so much that it’s just a reflex now
edit : the only way i think it can survive the export , unless it’s a format specifically Blender friendly, is if you :
-apply the subsurf.
-apply the edgesplit.
But then you end into larger amount of vertices (edge split doubling the vertices in edges it will affect) and lose the low poly model under the subsurf in case you want to keep it.
