I was thinking if it would be good to chose two types of SubSurf. One is the known, in the other, the surface would always touch the vertices. This is an example (it’s a only demonstration, no code): the cube subsurfaced from the left is the modified by the modifier as we know. The right cube, the new type.
What is this for?
Imagine you are made some low-poly characters for a game. The SubSurf is not recommended for BGE. But, in the future, the modifier will be really stable in BGE (I have hope), it won’t consume too much process and we’ll can use it as we want… the low-poly models with subsurf modifier wouldn’t be the same as we created them… the new surface wouldn’t touch the vertices. They would be too much softened (I hope you understand what I mean)
In this case, we also can add an option in the game to increase the resolution of the models if you have a fast computer.
In the modifier, there can be a “Catmull-Clark 2” new button, or can be a slider, from the modifier as we know to the “touch vertices” new option. Both would be usefull.
Are you talking about an interpolating subdivision scheme (as opposed to an approximating scheme like Catmull Clark?)
There has been discussion in the past of adding more subdivision types, such as Loop for better handling of triangle meshes, but no one has bothered much with it.
Nicholas, I don’t understand anything of maths, but I think you understood me. It would be the same face subdivision, with the difference in where the new surface supports (rests).
I posted this because I think it would be usefull especificaly for the use I wrote (BGE).
Some face details, for example, the nose, the eyelids or the ears, when they are in low poly and you use a SubSurf modifier, they loose details. It gets too much smoother.
This is another example: the face of Jim. To make the face looks nice with SubSurf, all the vertexs are exagerated. If you want the same model to use in low poly… you couldn’t. You see that it is horrible without the modifier, especially the eyes.
The difficulty is that you never get anything for free The reason for using approximating schemes (like Catmull-Clark) is that you can guarantee surface smoothness; with interpolating schemes you tend to get bulges, and in general they are just trickier to control. At least, that’s my understanding; I’ve never actually used an interpolating subdivision algorithm.
I remember a while back there was someone looking into adding Doo-Sabin subsurf but stopping due to the lack of n-gon support. Hopefully it’ll soon be possible to implement with the bmesh on it’s way.
That can already be done using edge crease. Looks to me like your really after a option that scales along the normals till it reaches the original vertexes.
Nicholas: Ok, I trust you, maybe you get bulges, but that is the intention of this type. It would be used only for some specific situations. But if you say that nobody would use it… then, it wouldn’t necessary to implement (I will ask some people who works in BGE, let’s see what their opinion is). Thanks for the fast reply.
N30N: The Doo-Sabin subsurf is a bit strange for me, I think it isn’t for organic models. I used to know about edge crease… but I’d forgotten them! Thanks for the memory refresh. Although, when the crease is at 100%, it seams like there is no modifier. And that’s not the point.
I’d definitely like to see doo-sabin subdivision added. I’ve played with topMod in the past in which it’s very useful for creating high-genus meshes with nice topology. I’m sure it would have lots of other interesting applications.
The reason for using approximating schemes (like Catmull-Clark) is that you can guarantee surface smoothness; with interpolating schemes you tend to get bulges, and in general they are just trickier to control. At least, that’s my understanding; I’ve never actually used an interpolating subdivision algorithm.