To preface upon delving deeper into the uses for this tool I have concluded firstly that I have only begun to scratch the surface of all the uses it has within an animation workflow, doubly so as I’ve only been using it for it’s static properties.

Since the Catmull Clark subdivision model was originally devised in 1978 up to modern times the standard for subdivision has been a straightforward linear b-spline division of each polygonal face into 4, the difference being averaged, eventually granting a smooth surface, generally by the 5th subdivision or so, as pictured below.

catmull-clark
The Standard Catmull Clark Subdivision Model

Pixar has over the years been updating this method in-house and has recently released their more contemporary methods for public development, with the intent of creating a standard across all 3D applications from inception of model to rendering stage. Known as the OpenSubdiv Project for the sake of universal adaptation, subdivision will now create smaller complex patches of subdivision polygons around key control vertices, with the rest of the mesh being made up of patches alternating a standard b-spline triangulation where possible. From this ‘transition’ patches will fill in the spaces in between with a dynamic arrangement of quads and tris to smooth the geometry while the patch itself remains the same shape as the standard b-spline segment. In this way only the primary vertex points take on an irregular (not a square) patch when subdividing, allowing the other patches to use minimal geometry to achieve a perfect curve in less iterations. The fact this rendering is now also primarily GPU based means it can be represented in the viewport while working much more easily as well. The video below explains this with graphic detail (vid starts 7:40).

 

I have begun this trimester to use OpenSubdiv as an alternative to turbosmooth for high poly modelling. In addition to the  benefits listed above the program also allows for much better edge-definition options via its crease definition without the need to add extra supporting edge-loops. Pictured below, the center model in each image is that on which the OpenSubdiv is acting upon to create the high-detail smooth on the left.

In as much I was able to create a suitable high-poly variant of a model from an object much more similar to my base mesh, keeping the poly count lower at both the low and high-poly stages.

For this model I set all of my creasing values manually, though for more complex models there is a ‘creaseset’ modifier available. This modifier allows you to quickly set values for groups of edges based upon angle values. The video below illustrates that this can be used for very complex objects, assigning crease values in a matter of minutes to thousands of edges.

Given that the place the improvements really shine is in regard to accurate pre-visualisation for render when animating I’m looking forward to applying what I’ve learned on the subject to character models in the coming months.

References

3dsmaxtrainer,. (2016). 3ds Max Tips and Tricks: Modeling Open Subdiv vs. Turbosmooth. Retrieved from https://www.youtube.com/watch?v=ktM61yLeXfM

Autodesk,. (2016). Exploring OpenSubdiv in Autodesk® 3ds Max® Extension 1 (part1). Retrieved from https://www.youtube.com/watch?v=ENA2FAF_PIc

Autodesk,. (2016). Meet the Experts: Pixar Animation Studios, The OpenSubdiv Project. Retrieved from https://www.youtube.com/watch?v=xFZazwvYc5o

Catmull Clark Subdivison. (2016). Retrieved from http://www.geocities.ws/jason_zxu/subdivision/pic/catmull-clark.gif

Introduction. (2016). Graphics.pixar.com. Retrieved 29 August 2016, from http://graphics.pixar.com/opensubdiv/docs/intro.html

Seymour, M. (2016). Fxguide.com. Retrieved 29 August 2016, from https://www.fxguide.com/featured/pixars-opensubdiv-v2-a-detailed-look/

Advertisements