SIGGRAPH 2013: A slice of Monster Pi

Posted By David Blumenfeld on July 30, 2013 08:57 am | Permalink
My third and final day of Siggraph came and went.  It was a pretty nice show this year, although as I mentioned before, it was smaller and definitely less energetic and enthusiastic overall.  I'm sure this is a combination of the economic downturn in this business coupled with the location.  I think it would be fair to say there also wasn't a ton in the way of new innovation being showcased.  Almost every year, there's at least one major theme (evidenced by the booths on the expo floor as well as the talks) in which some new development or tech topic sticks out, but this year really seemed to be more of the same thing with not much new.  I attended three talks during the day, the first being the production session for Life Of Pi, the second the State Of The VFX Industry discussion, and the final had some making of Monsters U and the Blue Umbrella information.  All were interesting, and I'll get into that shortly, but there was one topic I thought I would mention first.

I've been in the visual effects business since the mid 1990's, and while I've done my share of various types of work, it would be fair to say I spent most of the first decade working on feature films and shorts (with the occasional ride installation, music video, and special venue project thrown in).  Aside from around six commercials during that time, most of my work focused on large scale, longer term productions.  However, for the last six years, I have spent the bulk of my time engaged almost solely in television commercial visual effects, racking up easily over one hundred spots during that time.  As most everyone knows, commercial post production today is a different beast than it was a decade ago.  Every spot is made at a minimum of full HD resolution, with some larger format on occasion, and the type of effects necessary must be at the same caliber of those in feature films.  Of course, as schedules and budgets shrink, this work must usually be accomplished from start to finish in a few weeks at the most, using crews from small to miniscule.  This is what draws me to this work, and makes it fun and exciting.  While there is often a sacrifice on the r&d portion of the process due to time constraints, the ability to quickly figure out a solution, implement it, and create a number of shots in that short period is challenging and extremely rewarding.  The sheer volume of advertising that makes use of this type of work is increasing all the time, and due to this, there are a number of studios out there, both large and small, whose work focuses on commercial post production entirely. 

Additionally, because the hardware, software, and skillset used to produce this work is now identical to that of feature vfx, many artists cross over between the two types.  The reason I bring this up is that, as in the past, commercial visual effects is nearly entirely absent from SIGGRAPH in all forms.  There are no production talks on the making of these spots, there are no screenings (that I am aware of) of the top spots of the past year, and there is not even any mention of this sector of the workforce in the state of the visual effects industry talks.  I find it a great disservice to this large portion of practitioners and their work that this has been left out from the show, and I hope to see this added in the near future. 

As I watch my peers discuss their techniques on the films they are presenting, I see things that I fully understand and agree with, but have to forgo and "fake" in order to make the deadlines.  I am often able to match the quality in the result, but it is these types of techniques that I feel should be shared with the vfx community during these talks, as I think there is merit and benefit for those in attendance.  One reason commercial vfx houses continue to succeed (those that do, of course) is because they have found a way to keep their overhead low and the effects budgets in check by performing in a highly efficient manner.  This is an area that could be utilized by some of the larger scale film vfx studios so they can remain in business and operate at a more profitable margin, something that is of great concern recently and one of the many factors forcing these companies out of business or to look for financial relief elsewhere. 

Having worked at many of these larger studios in the past, I have witnessed firsthand a large amount of waste in regards to inefficient workflows, including too many layers in their vertical stratification, as well as development that, dare I say it, makes for a better SIGGRAPH paper than it does for a necessary step in the vfx production process.  Anyway, in summary, let's try and push for some commercial visual effects representation at the show, it is a large part of the community and would provide a greater amount of knowledge for the attending crowd.

For my first talk of the day, How To Bake A Pi, a panel of five supervisors from Rhythm & Hues spoke about some of the technical challenges and processes on the production, including the creation of photoreal animals, digital oceans and skies, artistic imagery (not just real but pretty), and working in stereo 3D.  Highlights included the problems they solved during production by constructing a practical in-ground 70x30x3 meter water tank complete with wave generators and special concrete "tetrapods" used to counter the effects of reflecting waves from the walls (called the bathtub effect) for greater realism of practically simulating an open body of water.  Houdini, Naiad, and other custom tools were used to realize the CG ocean extensions.  While not directly mentioned in the talk, I assume the underlying deformations of the water are based on the Tessendorf algorithms.  I have done some pipeline development at Brickyard using these algorithms and tools such as the Houdini Ocean Toolkit (ported to Maya) for this purpose, with great results.  For anyone looking to do open ocean simulation, this is a great place to start, and a simple web search will get you to some source code and precompiled tools you can use on your own for development purposes.  Back to the presentation, they further discussed the fx animation for whitecaps, mist, foam, churn, spray, and other water interaction, as well as splashes which were added as separate elements. 

Moving onto the discussion of their characters, they talked about muscle and skin development and bone simulations as well as the hair and fur systems for the animals.  Of note, the tiger had approximately ten million hairs, while the zebra was in the twenty million range.  I was a bit surprised by this amount, thinking it would be higher, as when at times I have had to do animal mug replacement, I tend to use approximately two million hairs for the front of the face alone.  They talked about some of the advances they made in rendering this fur, such as intelligent raytracing and importance sampling, coupled with partial hair transparency and ray occlusion, coupled with the use of subsurface scattering on the hairs themselves.  They spoke briefly about some of the crowd work as well, for shots of flying fish, meerkats, etc.  At the end, they presented a few metrics, the most interesting of which were a total of (if using a single processor) 1,633 years of total render time, with a peak disk usage of 260 terabytes.  In all, the talk was well done, informative, and the work they accomplished was beautiful.  I hope the talented team there is able to emerge from their financial woes and put all the artisans who created this back to work soon.

From here, I took my last walk around the expo floor, checking out a few technologies that I missed over the last few days, grabbed a bite to eat, saw a few more old friends, and then headed to the State Of The VFX Industry talk.  The talk itself started off with some good history of the industry and such, and then spoke to some of the major problems facing the business at this time.  I won't get into this too much as it can be a very polarizing subject and clearly open to interpretation, but I still encourage anyone working in this field to visit some of the websites they listed for more info, including:

I would encourage everyone to become educated on the subject matter so that any future discussion or participation can begin from an informed standpoint.

The last talk of the day was in the same location, and was about some various Pixar technology, including some areas of Monsters U and the Blue Umbrella short.  They spoke about the character rigging, crowd simulation and pipeline, vegetation such as trees, hedges, and grass population and simulation, and how this differed from that in Brave, as well as the rain and lighting/compositing pipeline for the Blue Umbrella.  The work was well done, though I honestly didn't find any of it groundbreaking or particularly different than before.  One thing I did take away, which I would like to look into and encourage the reader to do the same, is the NGPlant open source library, presumably a set of base code they took advantage of for their tree development.  As I may have mentioned before, I did some looking into SpeedTree software (a commercially available solution), which may be on my short list of upcoming software when I need it for a job.  I don't know if this uses a similar code base or not, but I will definitely check out this other library as well to see if I can glean any techniques or useful processes out of it.

In summary, I enjoyed the show and learned a couple new things, as well as found a few areas I will be researching more in-depth in the coming months. 

I found it interesting that the show seemed to be attended by a significant amount of people who didn't appear (at least at first glance) to actually be working in this field, at least in my opinion (one can never really know).  As mentioned above, I would really like to see more real-world examples of vfx production in areas besides large scale film production and university or tech industry research discussed, as there is definitely room for a greatly expanded set of classes, talks, and other show events that would be of interest to many in attendance.  I likely won't be going to next year's show in Vancouver, but hopefully I'll return to the show in two years if it's local to Los Angeles again.  It would be great to see it bigger and better than before, and I hope to see you there as well!