top of page

Behind the scenes of Gabriel Garzón-Montano’s music video “FruitFlies”

learn more

Thinking of our work as a gift can significantly change why and how we do it.

 

 

Go behind the scenes of Gabriel Garzón-Montano’s music video “FruitFlies” - to learn how a small passionate creative team managed a challenging project deadline with 500,000,000 particles using Houdini and Redshift.

The production team worked 16 hour days for two months to deliver the music video.  As the creative demands on the project increased, so did the project complexity.  With a narrowing delivery window, GridMarkets' Managed Services and Rush GPU offerings featured heavily addressing the complexity .  Read on to get the producer's perspective.

Director: 
Co-Director: 
DP: 
Producer: 
Special Effects:

Managed Services:

Simulating:

Rendering: 

 

Santiago Carrasquilla
Tighe Kellner
Oliver Lanzenberg
Jos Diaz
Saad Mosajee

GridMarkets

GridMarkets

GridMarkets

 

 

SUBMISSION STATISTICS:

  • 1500 machine hours

  • Houdini / Redshift

  • 60 CPUs and 120 GPUs

  • 500,000,000 particles

Santiago Carrasquilla: "One of the biggest challenges of this production was figuring out how long this video would take to finish.  We all had some idea of how ambitious our vision was considering the budget and the aggressive timeline. But it wasn’t until we were quite deep in the post production process that we realized just how iterative and tedious getting to a final product would really be. A solid 2 months of 16 hour days brings all sorts of physical and emotional challenges that are not always easy to navigate."

 

 

When we began production of the film began in December, we sketched a variety of options for how the animated light could look visually. Because the light was built entirely out of particles, a lot of this look development was dependent on particle count, scale, color, and post effects. The initial look of the light in the film tried to emphasize the individuality of particles more, and thus we opted for lower particle counts and larger particle sizes.

 

Toward the end of January, the team decided to shift the visual direction of the light heavily toward something more organic and fluid. This decision was made to try and emphasize the presence of the light as an individual character that interacts with Gabriel, the musician, in our video.

 

In order to achieve the more organic and fluid look of the light, it meant we had to change our entire technical pipeline to accommodate more computationally intense simulations. Our initial technical pipeline involved animating and rendering using Cinema 4D, X-particles and Octane. In our new pipeline, we shifted towards simulating in Houdini and rendering out of Redshift.

 

Prior to the look development switch everything was being animated and rendered on a single GPU tower.  This was possible because the particle counts were low and our productivity hadn’t been hindered. As we scaled the particle counts higher and higher, pushing the limits of what Houdini and Redshift could output on a single GPU tower, we began to run into various bottlenecks and technical obstacles. We considered turning to a render farm to aid with the production several times but opted not to due to the fast, iterative nature of our working process. We would often pool personal machines in the studio together to create local caching farms to process the simulation caches for the film as an interim solution.

FRUITFLIES LYRICS

A million pairs of feet, getting weak
We are all so tired of walking
But we can't find the way back home

Like fruitfiies in the breeze, ill at ease
We try so hard to fight it
But we can't change the way the wind blows
Can't find a way back home
We can't change the way the wind blows
No, we can't find the way back home

And while we try to sleep, forgotten dreams
They beg of us to count them
So we back and forth with the sheep
Our cloudy coats, they gleam
How we bleat
Afraid to up and wander
Let the wolf get you while you're alone
Can't find a way back home
Can't find a way back home
No, we can't find a way back home

A million pairs of feet, getting weak
We are oh so tired of walking
But we can't find the way back home

Can't find a way back home
Can't find a way back home
No, we can't find the way back home

 

 

In February, the team decided that some of the key story moments of the film (the first scene, the end scenes, and one scene in the middle) were important enough to the story that they would require a dramatically higher particle count to emphasize their presence. We used the very last scene of the film as a test case, in which the final render contains more than 500 million particles.

 

The scene was intense enough that we were frequently experiencing issues with application memory, simulation crashing, and production bottlenecks. It ended up taking us almost a full week to render a workable version of the scene, and we had little time to iterate or revise the shot because each render would take upwards of 40 continuous hours.  We lacked the CPU power to create a particle cache for the scene and would instead render it in one straight shot, relying on our machines GPU’s to do the heavy lifting.

 

Coming into March, we knew that the slow iteration speeds, memory/computational issues, and wait times would severely hinder us from scaling up the other key scenes of the film. Through referral from another artist, we came across the GridMarkets Managed Services solution.

 

Unlike traditional render farms, GridMarkets offered a flexible, collaborative service that would provide our production team with hands-on help from their VFX technicians. We selected GridMarkets because we knew it would be the best way to quickly scale up our particle counts reliably whilst allowing our team to focus on the artistic challenges of getting the film done in time.

Lessons learned:

  • We should continue to do things we don’t fully know how to do even if it hurts.

  • Finding like minded collaborators who are willing to go above and beyond purely for the love of making art is very precious and should be treated as such.

  • The balance between our ambition and real life circumstances is very difficult to find. The goal is certainly to find that balance so as to minimize ‘pain’ and increase flow and professionalism. However that balance should not be achieved by shrinking our ambitions but rather by making many mistakes and falling many times over as we dream bigger and bigger.  

  • Thinking of our work as a gift can significantly change why and how we do it.

  • A great way to assess the success of a piece of work such as this one is by remembering all the memories attached to the making of it.

  • When handing off Houdini files, always tell the person receiving the file exactly what version of Houdini the file was created with.

 

 

 

After finalizing the managed services agreement, GridMarkets immediately assigned two technicians, Adam and Steve, to our project. In addition to the two rendering technicians we were provided, Mark oversaw scheduling and progress of our shots as they were submitted to the farm and we waited for the caches and renders to finish.

 

We sent a total of 3 shots to GridMarkets, the simplest of which had 25 million particles, the most complex having hundreds of millions. The GridMarkets techs were able to quickly handle the simple shots we sent over, and for the more intensive shot, helped us optimize the scene to be rendered. Not only did they improve the overall optimization of our shot so that it could be cached and rendered quickly on their cloud, they also provided direct feedback about simulation caching strategies that our team was able to implement to improve the optimization of our local caching process. This was incredibly valuable for us, as we were able to accelerate the speed of both our cloud and local rendering for the production.

 

Toward the end of our production, the team ended up working long, late hours throughout the week and well into the weekend so that we could achieve the desired quality of animation within our tight deadline. Because GridMarkets has a global network of technicians, they were able to provide us updates and support on our renders at almost every hour of the day. This meant that we were able to come out of production meetings late at night, on weekends, or at regular times during the day and receive progress updates quickly. This knowledge was invaluable towards the scheduling and planning of the production.

 

Overall, the GridMarkets Managed Services option proved to be the perfect decision for our production. Our team found the GridMarkets technicians to be incredibly knowledgeable, easy to work with and very reliable. We never waited more than an hour to receive a progress update, and although we did run into different technical obstacles along the way, we were able to work with GridMarkets to resolve all issues until our shots were rendered exactly how we needed.

 

There’s always a grey area when one opts to use a render farm - about whether or not the shot will be returned looking exactly as intended in a timely manner. The GridMarkets managed services option is the answer to that grey area – it let us send incredibly complicated shots to be rendered on a short term deadline, and most importantly gave us peace of mind that we could focus on getting the film done on our end whilst the GridMarkets team helped us solve all rendering issues.

 

Our team can’t recommend GridMarkets' services enough!

bottom of page