Opensubdiv. What do you think?
polycounter lvl 4
Opensubdiv looks amazing to me but leaves some questions. Will it work in Zbrush or not since it uses creases for smoothing instead of control loops. Would working with creases help or hinder the modeling process? Is this another reasonable method for 3d modeling in the industry. I personally dont see many downsides to opensubdiv but I dont know everything and would like to hear other opinions on the matter.
Replies
But I've been using a similar system in Modo for some time now and it's pretty good, although doesn't quite give the crisp results that support loops give but it's a lot easier to manage and more lightweight geometry is always good.
It will be interesting to see what kind of things are done with it now, there is more to it than just variable edge weights. Hierarchical edits and feature adaptive subdivision could prove to be very interesting for real time applications.
https://www.youtube.com/watch?v=uogAzQoVdNU
Is the system in Modo actual openSubdiv? or is it some work-around the patent approximation? Because support loops should be completely unnecessary with opensubdiv
Not openSubDiv, no. As far as I understand it, it's an older version of the same kind of technology. Luxology traded some renderer tech to Pixar for Pixar's Subdivision Surface tech.
But that was several years ago, open subdiv looks to be better.
OpenSubdiv is the implementation of subdivision that also happens to use the same algorithm. OpenSubdiv is coming to blender soon as well
Edge creases are a godsend for hard surface modeling. Max users have been gimped all these years by not being able to use it.
I remember watching a Pixar modeling seminar and they used open subdiv for hard surface environments. The PC specs they used for the presentation was better than the average laptop but a gaming rig would work nicely.
That doesn't seem to be the case in practical use tho. What i have seen it creates a much denser mesh with evenly distributed geometry across the whole mesh, wheras with turbosmooth the geometry is located mostly around the supported edges. What's more is it seems you have to go a couple iterations above turbosmooth iterations to reach the same smoothness, hence polycounts are exploding even for rather simple meshes. Strictly talking about the smoothened meshes here btw. the non smoothened meshes are indeed lower poly.
Maybe i'm missing something here. Is there an option for automatic optimization or something? Would be good if someone who has tried the opensubdiv could chime in.
You are right about the easier modification of course. That's the whole point of opensubdiv. Tho in practical use that is a bit questionable aswell with how super dense meshes can slow down the PC to a halt and particularely how max has trouble to render subdivided meshes smoothly in edit mode.
I've been working for about 3-4 years with a manual chamfer workflow and the last 2 years with opensubd, so here are my pros and cons:
opensubd (or similar):
pros:
- cleaner basemesh, much easier to make tweaks and changes (no redoing support loops)
- cleaner subdivided mesh, much better for sculpting since faces are mostly square and the mesh is very evenly subdivided
- you can UV unwrap the basemesh and thus texture the highpoly quite easily, allowing for a more flexible texturing workflow (i.e. baking to first or third person weapons with different polycounts and UV layouts)
cons:
- heavier subdivided mesh, you need about one subdivision more for the same surface quality, since with chamfering you automatically add more resolution where it's needed -> on edges and bevels
- less control over bevel width, since they always depend on the size of the adjacent faces. This means you still need a few control loops to compensate for. If you try it at first, everything looks too blobby and smooth, but with some experience you can get very similar quality to manual chamfer
manual chamfer:
pros:
- best control over bevel width and appearance
- lighter subdivided meshes (better for viewport performance and baking)
cons:
- managing control loops is time consuming and makes tweaks and changes take much longer.
- Very hard to unwrap and UV due to the quite heavy distortion, plus having to manage control loops in the UV. Opensubdiv might alleviate the distortion part.
- Resulting mesh is often not sculpting friendly (lots of long, thin quads instead of everything being uniformly tesselated and square)
If you want to try what opensubD is like, just use ye olde double turbosmooth workflow, it's the same thing. Well, minus the UV smoothing and crease control part. lower TS modifier with 2 iterations and seperate by smoothing groups on, upper TS modifier with 1 iteration and separate off is equivalent to opensubD with 3 iterations and a fixed edge crease of 2 (for edges inbetween different smoothing groups). 3 iterations lower and 1 iterations upper means 4 iteration with crease 3, and so on...
Overall it's not strictly better, I would say it's faster and gives you mode flexibility, depends on how much value you put on this.
If you mean fillet width as in actual fillets or chamfers like nurbs tools or meshfusion, then no, opensubdiv doesn't have that.
-it's painful to watch at all these peoples who still use support loops when Marius Silaghi realsed his tools in mid 2012.
http://mariussilaghi.com/products/turbosmooth-pro
http://mariussilaghi.com/products/quad-chamfer-modifier
-it's painful to watch at Autodesk who needed a few YEARS to develop so important tool
I dunno if they are lazy or they just don't get the industry but definitely they have some problem with estimating "what's going on".
-but finally they developed it and it looks promising. I can't write more because I have this:
and a little gift: script for turning on\off the cage. It's usefull for workflow with open sub div \turbosmooth pro mod:
I'm hoping that the main difference will be speed, especially interaction speed.
I'm also not 100% sure if TS pro does the same exact uv smoothing you need to get the same result as you would when subdividing in mudbox. Autodesk opensubdiv certainly does, according to Neil Blevins.
Isn't performance the only thing that changes?
I am pretty sure you can use creases without open-subdiv and support loops with it.
Sometimes it's easier to crease than to add edgeloops. example
This inward extrusion in a low-tesselated sphere is a problem that's not easily solved with opensubdiv. Both chamfering and creasing mean that the corner points get displaced less than the surrounding low-density geo, making the ridge poke out, especially if it's a cube-shaped extrude like in the video.
There are reasons people don't use solely these tools. If I recall correctly, turbosmooth pro is just opensubdiv, which the pros and cons of using are being discussed in this thread. I'm not authority but I don't see it replacing control loops. Quad chamfer is a great tool which I use, but it is no replacement for edge loops. It has it's place though. And I don't see the need for that script when you can just toggle show end result with a built in hotkey?
Cool but did you notice that you didn't give any
An example how I use these tools:
I just selected edges and put some mods.
A clean high poly mesh fusion type boolean tool would be a cool advancement in Max's case beyond what's already been available (quad chamfer tools or manually adding control edges.)
I just don't see the point though, in my honest opinion it's still the fastest workflow for game/hard edge to create a "base mesh" that's 95% LP ready and just adding a quad chamfer modifier on top based on SG or crease values for your HP. That's been my workflow on FC3, Blood Dragon, and FC4. I do add some control edges to some complex areas the Quad chamfer modifier can't handle.
Here you go:
http://www.scriptspot.com/3ds-max/plugins/multi-mesher
mesh fusion for poor
ghaztehschmexeh,
Of course I'm not the smartest person on the planet and of course that everybody should use whatever they like but your arguments are funny so I trolled you a bit
I would say the biggest benefit is performance on skinned meshes. This is a Maya video, but I expect the Max perf to be similar. I don't think this will have a big impact on game development since we rarely do that.
Concerning workflow you are substituting manual control loops for crease management. So you should use what you are comfortable with.