Massive 3.0: A Real Crowd-Pleaser

The rat pack in Ratatouille, the armies of thousands in 300, the herds of animals in Evan Almighty, and before that, in 20 other feature films, all animated with Massive, might lead you to believe the software developed by Stephen Regelous originally for Lord of the Rings, is crowd simulation software.

‘We let people believe it’s a crowd system, but really it’s been about autonomous animation,’ says Regelous. ‘That scares some animators, but the whole point is to empower animators. They create the nuances of the character and then the character performs under their control.’

With that in mind, the company is announcing a multitude of new features in Massive 3.0, which they’ll introduce at SIGGRAPH. Foremost is a Windows-based version of Massive, FBX support for moving motion and skeletal data to and from 3D software programs, support for subdivision surfaces, dynamic hair and improved dynamics.

‘In 2.6, we could render subdivision surfaces, but you couldn’t see them in the user interface,’ says Regelous. ‘Now, we have interactive viewing. It was a bit tricky to implement, you might have 10,000 characters in a shot and the subdivision surfaces have to look smooth on every one, so we have different levels of smoothing.’

Dynamic guide hairs controlled by texture maps and rendered through RenderMan or Massive’s GPU-accelerated Velocity puts realistic Massive agents closer to camera. Improved dynamics adds complexity to their performances.

Massive agents typically perform a blend of predetermined snippets of animation created with keyframed or motion-captured cycles. The agents then modify those performances on the fly. ‘One character might pick up an object using a snippet of animation,’ says Regelous. ‘Then, because the agents can see, it can aim the object toward another character. The animator controls the nuances of the motion; Massive selects the actions and modifies them for the context. And now, we can have that animated performance drive rigid body dynamics. You get the best of both worlds: the subtlety from the animator and the realism of physics.’

For example, one agent might hit another that’s walking along. At the point of contact, Massive can switch on the physics without losing the performance. ‘Before, when the character hit the ground, that was it,’ Regelous says. ‘Now, Massive blends back into the animation and the character can stand up again.’

Massive 3.0 also includes new features to help people creating crowd simulation: enhanced lanes for directing such agents as cars along specific paths, layer simulations and real-time previsualization.

With layer simulations, animators can create passes for cloth, hair and rigid body simulations, for example, and then run any or all from within the user interface. With realtime previz, Regelous claims that animators can block out a crowd sim as fast as with a particle system.

While most animators use Massive most often for crowds, Regelous believes the software could also help animate hero and midground characters. ‘The agents can catch eyelines,’ he says. ‘It can solve the mid-ground problem.’