• Subscribe

The science behind Disney's magic

Speed read
  • High-performance computing enables incredible detail in CGI animated movies
  • Faster processing could improve the creative process and boost creativity
  • Senior research scientist at Disney calls for 'quantum leap' in HPC

Imagine a blank space before you. A flat, empty nothingness without a spark of life.

Now imagine you must transform it into a world: A vibrant landscape teeming with diverse organisms and populated by realistic characters with an important story to tell.

That’s the challenge faced by Disney’s team of artists, animators, and engineers each time they embark on what they hope will become the next Zootopia, Frozen, or Big Hero.

<strong>Devil in the details.</strong> Disney's appeal is due in part to their realistic animations, but the details we expect are outstripping the computational ability to render them. Courtesy Disney.

“We have to create everything you see,” says Rasmus Tamstorf, senior research scientist with Walt Disney Studios Animation Research.

That means every blade of grass, every grain of sand, every tuft of fur on the ears of the first bunny to join Zootopia’s police department. “Adding more details, adding complexity, adds appeal and believability to a world. It adds the visual quality that people appreciate.”

Though Walt Disney Studios have a long-standing reputation for hand-drawn animation and take pride in the sensibility and artistic language that derives from that legacy, their current films are all computer-generated. “There’s no way we could make the movies we do today without an immense amount of computing power,” says Tamstorf. “It simply wouldn’t be possible.”

Digital Disney

Computers are used in every stage of the creative process, starting with visual development in Photoshop. But once the basic elements are in place, that’s when computing power really comes into play.

“Once you start modeling assets in 3D, you have to come up with a look, an appearance for those models. It’s not enough to have just a shape,” says Tamstorf. The 3D models must be articulated so animators can control them, and place them into an environment — which is also created from scratch.<strong>Off the charts.</strong> Disney has seen an exponential increase in the computing hours needed to render their films. Soon, the layers of detail per frame will require a quantum leap in processing power to keep up. Courtesy Disney.

“Whether it’s a house or a rock or a grain of sand on the beach or a tree, it all has to be put together,” says Tamstorf. Only then can the animators start crafting the story they need to tell.

After that, other teams add clothing, hair, and soft tissue simulation to the 3D models, followed by environmental effects like water, fire, or smoke. The final steps of lighting and rendering demand the most computational power.  

The weight of magic

Disney’s animation teams are challenged by the both the large amount of data and by the problem of managing its complexity. Their storage requirements have ballooned to nearly 15 petabytes, while the compute hours needed have soared into the hundreds of millions.

Disney's HPC Innovations*

  • Frozen: Disney, with assistance from University of California Los Angeles researchers (and 4,000 computers), used advanced mathematics and physics to design breathtaking scenes.
  • Tangled: Unique techniques and tools create the impression of a painting — and the most expensive 3D movie of all time.
  • Big Hero 6: Disney's new rendering tool, Hyperion, required over 2,300 Linux workstations in four data centers, backed by a central storage system with five petabyte capacity.

These heavier computational demands ultimately enhance the Disney experience we've all come to expect, but also create bottlenecks for the animators and scientists behind the Disney magic.

“In any creative process, the ability to iterate quickly is really critical,” says Tamstorf. “The faster you can iterate, the faster you can fail — but you learn from your failure and eventually get to the right thing.”

In the new feature Moana, Disney has leveraged every bit of their computational power to create a rich Pacific world of ocean waves, swaying palms, and smoking volcanoes in which an island teen sets sail for adventure. 

One particularly complex shot in Moana would have required nearly 30 years to render were it not for multiple processors. Even with the advantage of HPC, the shot required 88 compute hours (nearly 3 days). For the whole of Moana, the iteration time per shot averaged about 12 hours, Tamstorf says.

There’s no way we could make the movies we do today without an immense amount of computing power. It simply wouldn't be possible. ~Rasmus Tamstorf

These reductions in turnaround time are an improvement, but to enhance the artists' productivity, further reductions are necessary — and this means strong scaling is on order.

Taking a fixed size problem and making it run faster by running on more and more compute cores is the type of scaling few HPC engineers are addressing, Tamstorf notes.

Disney's investment must be paying off, as Moana cleared $81M in its opening weekend. Despite this investment, Tamstorf sees a looming collision on the horizon, as the number of frames per movie will soon overtake the ability to process them in a timely manner. 

<strong>Riding a wave.</strong> Moana has set a new box office record, bringing in over $81 million in its opening weekend. It also has set the record for the most computing hours needed by a Disney film. Courtesy Disney.

“What we need are some quantum leaps [in processing power] where you really go from something that took days to something that takes minutes. But if we can get that leap, then I think we can fundamentally change some of those processes.”

Tamstorf isn’t waiting for his scaling needs to be solved on their own, nor is he expecting Silicon Valley to conjure the solution Disney desires.

Using the Edison supercomputer at the US Department of Energy's National Energy Research Scientific Computing Center (NERSC) in collaboration with researchers from the University of Illinois at Urbana-Champaign, Tamstorf was able to create perfect scaling up to 100 cores. These promising tests indicate strong scaling is possible on small and irregular problems.

Though more work needs to be done, Tamstorf found the results very encouraging. Once Disney makes the quantum leap Tamstorf is looking for, there will be no end to the new worlds and new adventures they will create for us.

From the blank space to the big screen and into our dreams, Disney's behind-the-scenes wizards won't stop working to make the magic we love to watch. 


Want to see more supercomputing magic? Follow the HPC crowd to Frankfurt next June.


*Courtesy International Data Corporation's ISC16 HPC Innovation Excellence Award Winners.

Join the conversation

Do you have story ideas or something to contribute?
Let us know!

Copyright © 2017 Science Node ™  |  Privacy Notice  |  Sitemap

Disclaimer: While Science Node ™ does its best to provide complete and up-to-date information, it does not warrant that the information is error-free and disclaims all liability with respect to results from the use of the information.

Republish

We encourage you to republish this article online and in print, it’s free under our creative commons attribution license, but please follow some simple guidelines:
  1. You have to credit our authors.
  2. You have to credit ScienceNode.org — where possible include our logo with a link back to the original article.
  3. You can simply run the first few lines of the article and then add: “Read the full article on ScienceNode.org” containing a link back to the original article.
  4. The easiest way to get the article on your site is to embed the code below.