Autodesk Project Geppetto is an early-stage research project that explores making it easy, fast, and fun to add crowds to 3ds Max scenes. Geppetto is a data-driven animation system that offers high level control of plausible human motion. Creating and editing Geppetto crowds is done using simple geometric shapes and paths. This free* technology preview of Project Geppetto comes with a set of animations for use in common settings like sidewalks, hallways, lobbies, and plazas. Architects, engineers, and designers can use Geppetto to enrich the context of their presentations with realistic human activity. Geppetto is the first crowd technology to offer both ease of use and high quality results.

Geppetto is part of a broader concept that we’re researching called "People Power." The basic idea is to try to assemble all the components you’d need to create, manage, and control large crowds of characters. Our research goal is to solve three big problems:

1. Creating believable motion easily (Geppetto)
2. Creating the look of any kind of human population or demographic and allowing for cultural influences
3. Creating an efficient framework for interacting with tens of thousands of such characters

What you're seeing with Geppetto is really just the beginning of #1 as we perform fundamental research on the other problems in parallel. At this time, we can’t make any commitment nor set any expectation that any of this will become a feature of any Autodesk product. This is simply a research effort that you’re being offered to explore with us.

Ocean of Motion

Autodesk has been researching the underlying technology behind Geppetto for over five years. The technique is more sophisticated than simple blending techniques that result in the awkward and implausible motions used in video games. Geppetto is based on a fundamentally new approach to how motion data is processed and applied to characters. Motion data from key frames or motion capture clips are synthesized in such a way that variations of the original performances can be interactively applied with a high degree of quality. The process of working with the data is akin to training your characters with performance repertoires. We’re calling this collection of motion that gets processed an "Ocean of Motion" to represent how different our approach is.

Importantly, the approach we’ve taken is not specific to human motions or to crowds. Given the right data, Geppetto could control dogs, snakes, dragons, cars, etc, and with a more directorial itinerary-based interface, Geppetto could be used to block in individual "hero" animation. Because the technology is data-driven its capabilities are limited only by the amount and kind of motion data it has access to.

Geppetto technology can currently solve some of the following problems:

• Path following: Real-time steering of physically correct walking and running motion that follows a specified path.
• Agile responses: Real-time triggering of physically believable agile motions, such as quick turns. This is required for collision avoidance and navigation.
• Object interaction: Seamless and natural real-time interaction with objects in the environment, such as sitting in chairs, or stepping up to climb stairs.
• Intelligent "human-like" dynamic obstacle avoidance: The perception of potential collisions and subsequent evasive actions must mimic the response times and behavior of real human beings.
• Intuitive crowd orchestration: New methods will be introduced that allow artists to directly control the flow and interaction of traffic patterns. Ease of use plays a major role in the design of the intended workflow. Characters are orchestrated in an intuitive high-level fashion through the manipulation of flow patterns, goals, and designated behaviors. The tools are geared toward controls that are fun to use, and accessible to "non-animators" but not at the expense of serious artistic control.

Limitations

Given the enormous range and intelligence of how humans move in crowds, a guiding principle of the Geppetto development goals is to gradually increase the depth of complexity from a framework that begins with a restricted set of behaviors. The Project Geppetto technology preview is intentionally bounded in scope in order to understand, test, and assess the perceptual "plausibility" of results with relatively simple, manageable crowd scenarios before adding in more complicated motions, skills, and interactions. There are some major limitations in this first manifestation of Geppetto, those details can be found in the User Guide. You’re being asked to evaluate a subset of what we know is needed, not so much to tell us that the known limitations are indeed limitations.

Autodesk Labs Goals

Our limited goals for this Labs release is to test our early concepts with a broad range of users:

1. Is the Geppetto motion believable? Did we succeed in creating life-like behavior for a limited range of cases?
2. Are the methods for setting up and configuring the crowds useful, limiting or too difficult?
3. Do you think we’re headed in a generally useful direction?

As you can see, things like the representation of the character is not a goal for this initial release because what we have is not what we want. Also, it isn’t a goal to see if you can use this initial release with all terrain, buildings or structures - we know it only works with flat environments.

Future

We can't comment directly on our plans for this research, but we’d like to see this technology become more flexible so that it can be used in more arbitrary ways, whether for analysis or entertainment purposes. Project Geppetto is just a stepping stone to much more interesting applications of this technology. The technology preview executable expires at the end of the technology preview based on a time-bomb date that has been set for July 1, 2012.