Obviously, this is a complex undertaking and many issues were bound to surface.

Problems

Modeling directly in VRML is very limiting, so we used a 3D package, trueSpace 4.1, to model and export to VRML. This allowed us additional capabilities, but increased the difficulty of implementation. We dealt with several compatibility issues exporting our models into VRML. Also, we found that the SGIs won't handle a model unless it is pre-triangulated, something which we were able to accomplish before exporting the model, but only after we knew we had to. We found it was necessary to limit the polygon count of the model because the interactivity of our project was slowing down even a 3000 polygon model quite noticeably. The other unexplained problem was that the NT machines ran the model rather well and the SGIs would not do as well in many instances. It seems odd that a more powerful machine would run the model less effectively, especially considering that the polygon count was so low that it should not have been an issue for either machine. It may have something to do with the way our 3D package was exporting the model, but we still have no definite explanation for this. There were difficulties in capturing the mood of our original design ideas (see Structure) because VRML handles things very differently than most 3D packages. Lighting, particularly, was a challenge, as things tend to be either bright and washed out or dark and invisible in VRML. Although we found a happy medium, it is still far from ideal.

Shortcomings

The model still runs relatively slow, even on an Indigo2 R10K. Motion of the player interface is a bit choppy at times, with the faders not giving a smooth continuous audio transition at all times. Mostly, though, we're disappointed that, despite getting the polygon count down from over 50,000 in the original 3D models to just under 3000, things will only run acceptably on one specific computer. Even other Indigos in the U of M VR lab were behaving sluggishly. NT boxes have run the model decently, but don't handle the multiple audio files as well as the SGIs.

Having said all of this, click the "view the VRML model" graphic below (or on the home page) to check out the working versions of Ossia.