You’ve started your own studio. You secure your first feature film. Everything is great. So you invest in the people and technology like Redshift to bring it all together. But there are unforeseen delays. And you don’t have the resource to make up for lost time when you come to the final renders.
Find out how nGenious used YellowDog to speed up production and deliver their feature film made with Redshift on time in this Cloud Render Case Study.
NGenious describe themselves as ‘young, excited…and reinventing the way we make independent animated feature films’. Ngenious used Maya, Redshift, and YellowDog’s GPU Cloud Rendering to deliver their inaugural feature animation: ‘Troll: The Tale of a Tail.’
We sat down with nGenious Co-Founder Eric Gervais-Despres to find out more about their feature film production. We also explore how the studio plans to compete with some of the biggest and best animation studios in the world in the future.
Tell us a little about ‘Troll: The Tale of a Tail’.
The movie is about trolls…and not those toy trolls with the pink hair and glitter and lasers. These trolls are quite different. The film is a family film. There are farts and burps for young children but there are jokes for older children and adults too.
What software was used in the pipeline?
Most of our artists worked in Maya and mainly Linux. Software like Mari or Photoshop doesn’t work well with Linux so we’d switch to Mac for texturing for example and then switch back to Linux after. We used Redshift for the GPU rendering and YellowDog for the cloud rendering that our GPU farm didn’t cope with. Finally, we used Nuke for compositing.
Why GPU rendering?
Using a GPU render pipeline was a decision that we made very early on.
In large studios that I’d worked with in the past, we were aware of Redshift as far back as 2015/2016. Well established studios have usually invested so heavily in CPU render farms and pipeline tools that it doesn’t make business sense to switch to a GPU render pipeline. That type of studio has traditionally stuck with engines like Arnold for cost reasons as well as those tools being proven for quality. Why change?!
Why was nGenious different?
We had no existing rigs or farms or infrastructure when we started nGenious. Nothing other than ourselves. With the clean slate, we were free to decide our path. I’m happy and proud with the shots that we delivered with our Redshift pipeline.
To my knowledge, we were the first studio to embark on production of a full feature film using Redshift for all of the rendering. We had 1,700 shots. We had 82 minutes of animation which we had to render in stereo. We hit the rendering hard: we had jungles and plants in a lot of shots. The renders were demanding. It was a heavy load to throw at Redshift and to throw at our GPU farm.
In mid-2016, we benchmarked other GPU engines like Octane. At that time, Redshift was the one that was the closest to what we wanted in terms of speed and look. The benchmarking process was thorough and the choice of Redshift was based on data that we collected from our studies.
We weren’t rendering final shots in 2016 but every time we built an asset it was creating a Redshift proxy. From the first step, from the first 3D model, Redshift was at the heart of our pipeline. The decision we made early on therefore couldn’t change.
How much GPU power did you have for the Film?
Locally, we bought what we calculated to be enough render power plus contingency. For our local GPU farm, we bought 20 blades with 4 high powered GPU cards each.
On top of regular production delays like story changes or new iterations, we had the added challenge of building the pipeline at the same time as making the film. For example, if we were making a tool for lighting and the release of that for production was delayed, it would also push the entire schedule.
When Nadja Leblanc, our production manager, recalculated our render delivery capability toward the end of the schedule, we knew that we wouldn’t make the immovable final deadlines.
What options did you have for more GPU render power?
Some people said that we should buy more GPU cards but there were two main problems. Firstly, the cost was changing hourly because cryptocurrency activity was volatile at the time. Secondly, the availability was low so there was no guarantee that what we ordered would get delivered in time.
We looked at dozens of render farms or cloud options. Honestly, there aren’t a lot of GPU render solutions in the world that could accommodate our needs. We found some options in China and India but many of them were offering 5 or 10 blades with no real integration benefits and very little security.
My partner Benoit Parent, who is also the CTO of the studio, made me realise: “Well, we have more power than these farms! These aren’t solutions for us.”
I emailed a private industry group asking for render farm suggestions. Someone recommended YellowDog after they had used their Redshift cloud render farm. We did more research and found out that YellowDog had access to hundreds of GPU machines. They could source more very quickly if we needed them to. Nobody else could give us that amount of GPU power with proven credentials in animation rendering; that gave us confidence they were the best solution.
What were YellowDog’s biggest impacts?
In the end, we were using hundreds of their machines with dual Nvidia Tesla P100s. YellowDog has incredible GPU cloud rendering capabilities. But Benoit was always keen to point out that ‘power alone is useless.’ You must find a way to use it efficiently.
The YellowDog team worked with Pierre-Luc Boily, our Pipeline Supervisor. A simple and smooth integration with our customized pipeline was implemented. We were highly appreciative that YellowDog developed new features or improvements too – some of these were specific to our workflow.
Did you monitor renders closely?
The YellowDog Platform was great because it meant that we could see exactly what rendering was in progress at any point. If something wasn’t right, Pierre-Luc could change something quickly himself or ask YellowDog’s render wrangle team to check on something.
The expertise and the partnership that we developed meant that we could harness YellowDog’s power efficiently. The combination of raw render power and their dedicated support gave us lighting speed; speed was invaluable to us at that point.
The film was delivered on time just as we wanted. What more can I say but thank you to YellowDog?!
How did the studio start?
I was working in Montreal as a Technical Director and consulting for producers. One of the producers I had been working with was heavily involved in an animated feature.
The producer said one day: “You know Eric, it is too bad you don’t have your own studio, you would be the perfect person to make this film happen.”
I’ve been in working in animation and visual effects for 25 years. I’ve had the privilege to work at Softimage, Walt Disney Feature Animation, and Warner Brothers. I’ve been involved in productions like Tarzan, Fantastic Four, 300, The Wild, and Gnomeo and Juliet.
I said back to the producer: “Maybe it is about time”.
With my business partner Benoit, we started the studio. We were buoyed by the confidence of knowing that we had our first animated feature secured.
What were the first challenges when starting on the film?
We were very lucky to have the project green lit to work towards. But that put us under considerable pressure to do things quickly and balance lots of fundamental processes. Not many studios have been in the situation where they need to be configuring electrics for workstations whilst configuring the pipeline or building the actual walls of the studio whilst building 3D assets for the film.
What is the culture of nGenious?
Benoit and I wanted to play to our strengths and experience. My background is developing technology and pipelines; fundamentally I’m an artist. For a time at Disney, I was Technical Lead for Layouts and Backgrounds for the entire studio across Burbank, LA, and Paris. Benoit was at Autodesk for over 20 years . He’s a genius in hardware strategy, setup, and technology. He is an inventor. For our own local render farm, Benoit created an automatic trap attached to an electronic thermometer. This enabled us to capture the cold air of our Montreal winters to cool down our blades and reduce electricity costs compared to other studios who may have to use traditional air cooling.
How did you try to stand out?
The experience I have from working with different pipelines on ambitious, complex, global projects at places like Disney is precious to me. Benoit’s technical experience is invaluable. I felt that we had the opportunity to combine the advantages of a fresh, nimble, small team with the ambition and knowledge of large studios.
There are many incredible studios in Montreal like Cinesite, Framestore, MPC, and others. There are many incredible productions being made. There are many jobs. But, there are not enough people to do them all. We knew that we had to stand out from day one.
For recruiting talent, we focused on creating a healthier work-life balance than other studios . We tried to foster an original culture to attract the people that we needed.
Is ‘different’ important when producing films?
How can technology make our lives, our films, our artists better? They are the things that people really want when they’re looking for their next job in animation. That’s my opinion anyway.
But the advantage is not only for our team or for our internal culture. Being big on technology, like YellowDog, and innovating pipeline solutions means that we can make movies way faster than other studios. Our mission is to make movies that cost $15 million look like they cost $60 million.
What does the future hold for nGenious?
We will continue to focus on animated features for now. In the future, I want to make movies that look truly different.
The truth is that a lot of animated features look the same. They’re beautiful and well-crafted but they’re all created with the same sort of workflow and they all end up hitting the big screen and often becoming indistinguishable in their final look.
I’ve seen concept art for a CGI production that is truly wonderful: captivating and artistic. And then it gets put through the pipeline and comes out and looks like every other animated CGI film that’s ever hit the big screen.
What is inspiring your vision?
The new Spiderman animated feature bucks the trend. That’s the sort of approach I have in mind for nGenious: make better films in a better way. Make them different.
I believe that right now, most films look like Picasso. Picasso is a great artist; many of the films we see are great too. There are differences in each of Picasso’s paintings. There are differences in the films we see every year. But, in the end, they are all Picasso. I don’t think that we have fully explored the space for more art in cinema, for different types of artists on the big screen. Picasso is wonderful but there is room for Monet or Dali or Hockney too.
‘Troll: The Tale of a Tail’ is currently on limited release and is set for global distribution in 2019.
People who enjoyed reading this, like this YellowDog case study with M2Animation, who used YellowDog for their Redshift renders too.