Maya+zSpace & Deja Vu

Description:

zSpace was a very impressive system. It had many attributes that set itself apart from other solutions. With sensors tracking the zSpace glasses, users could tilt their head to look around an object as zSpace dynamically displayed the correct perspective. Using the 6DoF zSpace stylus, which is held in midair like a pen, users could rotate their wrist naturally to pick up and examine objects. It lent itself to jaw-dropping demos.

As much as our early demos were useful experiments, they were also toys. As a company, zSpace had to make something to prove our applicability to real-world problems. We had already identified that the creators of film/TV/video games were a possible target market and zSpace had useful connections inside ILM and Autodesk.

Working with Autodesk and ILM, we chose to integrate zSpace into Autodesk Maya. Maya’s 3d design use cases and technology stack made it a great target for our goals of demonstrating tangible user enhancements to an industry-standard application. Specifically, we believed Maya+zSpace would be more intuitive and more immersive for modeling workflows and that users would produce higher-quality content faster than Maya alone.

By conducting workflow analyses of routine use cases for 3d modeling, I created a list of the most commonly-used Maya tools for 3d design: Selection, Move, Rotate, and Scale. After I designed the zSpace-equivalents of these tools and they were implemented in Maya, I conducted a usability study of the Maya+zSpace combination using expert Maya teachers and practitioners as subjects.

The study’s positive findings were that subjects loved the viewing experience provided by zSpace. Direct quotes included: “That is beyond bitchin.”, “This is very fun; very visceral.”, and "Love looking at this thing. It's neat!”. However, the study’s negative findings were that subjects needed exact pixel, angle, and scaling percentage control over their assets that zSpace’s tools did not provide. For example, creating a wedge shape that was 10cm long with a specific slope was harder in Maya+zSpace than it was in Maya alone.

Based on these studies, zSpace hit its goals of being more intuitive and immersive than Maya alone but failed to empower users to model more quickly. From these findings, the business decision was to position the Maya+zSpace offering as a quality assurance review tool for models and animations rather than a full modeling environment.

For internal stakeholders, I highlighted a key negative finding from the study as an overarching issue we had to resolve regardless of the path forward. Since Maya was originally designed for mouse+keyboard interaction, all the menus and buttons were very small hit targets and zSpace users had a very challenging time selecting items in these ‘legacy’ 2d UIs. As one usability subject put it, "I really have to aim.”

The problem was that, as the user pressed down on a stylus’ button, the pressure introduced angular displacement of the virtual stylus beam. By the time the button’s electronics registered contact, the beam was no longer pointing at the intended target. Since this is an angular effect, it is exacerbated by how far into the scene the target item was located. Working with an engineer, we gathered usage and timing data for this specific problem. This was a complicated, multivariate problem and we were very concerned that the solution would be complex and brittle. However, in analyzing log data, we discovered that the time between hovering an intended selection target, starting to press the button, and the electronic contact taking place was very deterministic. The solution, named Deja Vu, was to create a very small cache of transform information for the virtual stylus beam and, when the system recognized a button contact, the system would use the beam’s transform from ⅓ seconds before the contact occurred. Deja Vu turned out to be invisible to users and so robust that it became the system’s default behavior for all selection operations.