The difference in performance is really impressive .
It’s not clear from the video, but is that the new and sexy feature-adaptive subdivision, or the old and busted uniform subdivision?
Also, the opensubdiv demos showed some impressive view-adaptive tesselation and displacement. Is that planned as well?
I cannot wait. 
Amazing, will It have crease edges support?
Amazing, good to see an official demo. Kudos to Sergey.
GPU tesselation is possible only if SubSurf modifier is last in the stack
Is this a limitation to be solved? What about non topology affecting modifiers? Having fully displaced mesh (at high fps) in viewport is one of the strongest point of it.
What if they added the skin displace map via modifier after SubSurf? Would gpu acceleration been stopped? (and fps dropped?)…lot of questions, it’s great to see it coming, but i’m missing the big picture and workflow they have in mind.
So… no displacement after subdivision? If so, that needs to be fixed ASAP. One of the big reasons for OSD is the ability to see where high res displaced meshes are in real time for contact animations and stuff. Disabling displacement basically just gives eye candy without much use for production characters.
yes I think that was one of the cool things I saw in pixar’s presentation videos… hope they can fix this
@m9105826: from 0.2fps to 40, it’s an eye-candy of monumental proportions…
I’d like to know more about that as well.
Sergey says it is only beneficial for animations, but in the meantime, his viewport performance is huge, even when posing and not only when playing back animation. oO
And is it another modifier ? Or will it replace our current one ?
It’s be great if Sergey could come over and explain us the limitations and future implementations of this !
Make sense don’t have support for a displacement modifier after the OpenSubDiv, but i think OpenSubDiv have it own way to make Displacement so what we need is to expose inside the OpenSubDiv Modifier
The displacement would need to be able to preserve the normals and the tangent space across the tesselated surface anyway (to support normal maps), so the current displacement modifier wasn’t really an option anyway, was it? It’s ok for stuff like terrain heightmaps, but I never saw it as a way to do details.
It seems that the open subdiv will only be for the animators to animate easily, not for modeling, so just a modifier to place to te top of the stack.
Same for alembic, only for the cache.
I hope it will not be that.
I am not sure but I think it is vector displacement maps. Displace modifier have a XYZ->RGB mode that maybe could handle that kind of maps.
I did not see people showing great example about that. So I am not sure that it is a satisfying way to handle that.
I am only using Blender. And Blender is not able to generate that kind of maps or Ptex layers.
So , it would make sense for people working with Zbrush, Mudbox but not for people using Blender Sculpting.
Nazgul is talking about Multires workflow has the most used workflow.
But blender artists are in love with Dyntopo not only for basic shaping but also for details creation.
So, I think that High res/Low res workflow has a future.
People were amazed by Ptex for painting “proof of concept GSOC”.
But it is probably a more longterm ToDo than fluent animation of models subdivided to several millions of polys in viewport.
First of all I would like to say that I am totally ignorant, technically speaking, and that the speed boost saw in the video is already a great thing.
With that said, I wasn’t talking about sculpting, but displacement. I remember a video of pixar, where they were said that the really nice thing about this technology is being able to animate the displaced model (so, maximum subdivision level and displacement) directly in the viewport and in realtime (or however reasonable speed). Again, the speed gain from the usage of openSub is awesome, but being able to see where the model surface will actually be during the render while animating it in the viewport would be even more awesome.
I think it’s safe to assume that this is only an initial implementation and I hope they don’t stop there. Opensubdiv, if fully implemented would completely change our workflows for modeling, animation and rendering. All for the better.
I don’t think its meant for modeling even when fully implemented
It depends on how fast the FAR can be computed. There would probably be a little freeze with topological edits, but it should otherwise work quite well. Seeing as our current subsurfs start choking at around 5k poly cages and 3 or 4 iterations (and you need to model with 3 or 4 iterations to take advantage of opensubsurf’s creasing feature), opensubdiv might actually be more usable. The important word here is might. There’s no real way to tell other than to try.
@Piotr Adamowicz thanks
In an attempt to educate my ass further I found this link http://www.fxguide.com/featured/pixars-opensubdiv-v2-a-detailed-look/ I thought some here might find it helpful.
I really hope we get both, cc and opensubdivs. Otherwise (if you look at the current limitations of opensubdivs) it would actually be a downgrade modeling wise.
I am just looking at the benefits of this OpenSubDiv, but as a modeler of mostly (lately) very curvy subd modern furniture pieces, much work on the model is done in the “streamlining” phase: not often adding/changing geometry but pushing (grab/rotate/scale) verts and edgeloops, often using proportional editing and all this on highly subdivided meshes (and the subd could easily be the last modifier in the stack here). Wouldnt OpenSubDiv be a huge speed boost in such modeling situations? I might be wrong but seeing the posing part of the video, Im assuming Im not.
Watching that video, I really like the crease sets. Seems easier to manage than how I currently use creases in Blender.
I realize this isn’t necessarily an OSD function, but maybe rather how Maya does things.