Superscape's SeV

Dateline: 4/19/00

Superscape a company with a long history of work in 3D is now jumping into the shark infested E-Commerce waters. SeV (Superscape e-Visualizer) is the new flagship visualizer. It's a very interesting technical piece of work. The business model and approach seems to be straight consulting and services type work. First let's look a little at the technology.

SeV requires a plug-in which did install quite seamlessly. (Yes you're thrilled already, another plug-in.) SeV uses it's own software renderer and isn't dependent on DirectX or OpenGL or whatever other renderer of the month is present on your machine. The software renderer seems quite impressive and as stated on the Superscape Web site:

"In addition to displaying lit smooth-shaded objects in true color, and transparent or semi-opaque objects, the browser supports advanced features, such as bump maps..."

"The software renderer supports dithering for screens running below 24-bit depth, and antialiasing in order to give the best quality image on screen."

SeV glass
Glass vases with interesting lighting. Image Courtesy Superscape.

Of course this does leave one wondering about hardware acceleration and those sorts of details. The Superscape renderer does however offer some striking advantages. Key among these is the use of subdivision surfaces.

Subdivision surface demonstration
Subdivision surface demonstration.

Smooth curved surfaces can be represented extremely compactly. Subdivision surfaces provide a clean way of managing levels-of-detail (LOD) as they can be generated on the fly. In addition subdivision surfaces offer other advantages, as explained by Paul Beardow VP of Research and Planning:

..they are more powerful than simple LOD substitution in that you can continue to generate detail as required, i.e. when zooming in, rather than switching between pre-existing models and running out of detail at the highest resolution downloaded. This is why Pixar liked them so much, you can model something once and not have to create new versions if you need more detail. On the other hand, you can easily ramp back resolution as required for lower spec machines. Traditional LOD is useful for substituting complex branches of the scene tree with simpler ones and this works nicely in conjunction with subdivision, the best of both worlds if you will.

Of course the biggest advantage of subdivision is visual quality - because they are generated, the lighting normals are exact and not a result of an approximation from looking at crease angles. This means that surfaces are very smooth at all resolutions and under animation. Not only that, but you only need to animate the simple starting mesh to animate the final generated one, so the number of points and hence the data needed to be transmitted is much smaller. Overall file size is much smaller than other methods as well because you only need a starting mesh. Plus of course instant gratification for the user at the final resolution rather than having to wait for the scene to build up.

The actual quantity of data transmitted is tiny and better than any other Web3D technology I'm aware of, including VRML or Metastream. The performance for the small objects on the Superscape web site is also terrific.

In addition to the surfaces themselves the animation of objects and characters can be produced with dramatically smaller bandwidth requirements. The animation of a subdivision surface can be controlled by manipulating a control mesh. The animation simply records changes to the mesh which then moves the entire geometric object. Since the geometric object is stored locally on the client only the control mesh changes must be downloaded a far less demanding task. Let's look a little at the Web integration and authoring (or lack thereof) issues.

Next page Integration >Page 1,2