Getting Realistic
Turlif Vilbrandt
November 28, 2012

Getting Realistic

Saving the World from Polygons: Moving toward the complete modeling of real things.
CG artists and designers know very well the limitations and tediousness of modeling with polygons. Mesh models tend to have all kinds of problems, such as cracks, holes, and self-intersections. This is due to a disconnect between the representation of the real world and the attempts by the modeling software to represent real, volumetric, complex, and “messy” objects by just using surfaces. How does a polygon modeler create digital dirt and goo? 

Currently, the modeling process for the user entails twisting, bending, sub-dividing, painting, and gluing together pieces of virtual paper to make an illusion of a real thing. This is acceptable for producing images and movies, but the objects that are modeled bear little true resemblance to real objects; instead, the representation is “skin deep.”

The emergence of 3D printing and especially a trend to multi-material printing is challenging existing modeling software and the geometric kernels at their core. It is radically changing the drive and needs of 3D modeling platforms and forcing a revaluation of existing tools. Simply speaking, bent and glued surfaces are not good enough to represent physical tangible things. This becomes even more of a serious problem when bio and medical modeling and bio-printing tissues are involved.

It becomes necessary to model using real volumes, whereby an object is thought of in its entirety with its surface and internal distribution of material, density, color, and even temperature and other physical properties. One can say that volume modeling existed for some time in the form of voxels (3D raster images). However, this is a fairly limited modeling approach due to its discrete nature and limitations imposed by raster sizes.

Uformia AS in Norway ( develops the new-generation 3D volume modeling and fabrication platform Symvol (currently partly available as a plug-in to Rhino), which is built from the ground up using real or true volume modelling, unlike the traditional industrial approaches based on meshes, parametric surfaces, or voxels. Not only can this platform model 3D geometry, but by using the same uniform process, it can model any number of properties of an object, such as materials, including the complex and dynamic blending of properties. This means that when designing objects with Symvol, the user is creating realistic models that are understood as true solid volume with materials and attributes inside. Just as in real life, there is no such thing as an object without thickness or substance. The users can think of clay or metal when modeling volumetrically. Whatever they do to the object, it remains a piece of malleable material at all times, unlike in traditional 3D modeling where objects can have surface issues and are empty inside by definition. This is achieved through the representation of every volume object as real mathematical functions. 

Because Symvol is using real 3D volumes, there are a lot of things that users can now do that would be otherwise impossible or very difficult in existing software. Perhaps, they want to design a toy that looks like an elephant but with features of a hedgehog, or they want to take a complex natural shape like a diatom and combine it with their own design. They can simply take two models or even scan data and combine them though any number of operations, including dynamic, automatic morphing between shapes, without worrying about how or if it is possible. 

While 3D printing is a revolutionary technology, it can be a painful process going from a 3D model to a physically fabricated object, and using existing design tools, it is largely a one-directional process. Users have to step through a set of stages, including surface mesh generation (for instance, from NURBS, Sub-D, T-Splines, and so forth), often at resolutions below the capability of the hardware (or modeling within meshes directly, which has its own limitations); analysis of the model to look for surface errors; automatic repairing or "stitching,"; creation of support structure for many machines; slicing the model into layers, which can generate errors for complex meshes; and, finally, 3D printing of the object (taking hours or days based on the process used).  

If any of these steps fail or the user wants to make even small adjustments to the model, the user must start over and, at minimum, remodel parts of the object that have failed.

Symvol eliminates this burden and allows users and 3D printers to fabricate directly without the need for this complex, multistage, fixed process. However, it needs to be mentioned that many 3D printers will currently not allow the direct delivery of sliced data, although they could (this is changing as better standards than STL meshes are currently being developed). Objects created with Symvol are always "watertight," and users can directly output accurate layers at any resolution no matter how complex the model —making the elusive, easy to use  "3D print button" a reality.

Teabunny - Mathieu Sanchez (Bournemouth University)

The best way to understand what real volume modeling means is to look at some examples of the change in the approach and process. The Teabunny presented here is a combination of two iconic models in CG: the Stanford Bunny and the Utah Teapot. Both polygonal models are automatically converted to functional volumes. This conversion can even make use of bad or broken mesh data. At this point, they are first-class volumes than can be modeled like any native volume in the system. After converting them to volumes, it is very simple for a user to create a union relationship between the two models that can include a blending property. 

Once this union is created, the objects can be modified and will still maintain the union and blending properties applied, which can also be changed at any moment. This makes for a very dynamic, painless, and playful modeling experience. The teapot can simply be moved to another part of the bunny, and the two maintain a union and any blending that was applied, while keeping the resulting model watertight at all times. This example, while simple, illustrates the painless incorporation of existing data and of some of the fundamental changes in store for the 3D design process. 

Using real volumes puts a focus on what you want to model and not how it is being modelled, allowing many more individuals to participate in 3D design and printing as well as the creation of new easy to use interfaces.

While real volumes can make 3D modeling easier, it also provides new capabilities and tools to the designer that have typically been considered hard or impossible. These new features are allowing the application of 3D design and manufacturing to fields that have traditionally not been considered, such as the design of living tissue. For example, internal attributes and complex internal geometry, such as cellular structures, which is a becoming a critical part of modeling for 3D printing, can also easily be applied to any model. 

Minotaur Head with Lamella - Neri Oxman in collaboration with W. Craig Carter (MIT), Joe Hicklin (The Mathworks) and Turlif Vilbrandt (Uformia)

Symvol was recently used to design and create one of the first truly heterogeneous objects with multiple physical properties through explicitly controlled fabrication. The object incorporated an opaque hard material and a soft rubber-like material into an imaginary helmet as a proof of concept and was recently on display at the Centre Pompidou in Paris. It was designed using bio-medical CT data to inform the shape and internal properties of the object based on the relationship between bone and soft tissue.

The future of 3D modeling will be much more about physical objects rather than rendered images. It will be about real volume modelling, not graphics, and it will put a premium on making sure we can capture the true nature and construction of real objects, not just an abstract set of plain drawings or a skin-deep illusion. 

Founder and chief technical officer of Uformia, Turlif Vilbrandt has a long history of researching methods and processes to exactly describe (computationally capture) the complexity and quality of natural and real things.