Koen Vroeijenstijn worked in the video game at Guerrilla Games before coming to the U.S. and joining DreamWorks Animation. He worked on projects such as KUNG FU PANDA, MEGAMIND or RISE OF THE GUARDIANS. He then joined Image Engine to work on ELYSIUM.
What is your background?
I started university in Industrial Design but after two years switched to Physics. At the same time a friend and I started a small company in Amsterdam. We started out with architectural visualization. We soon expanded into visual effects for commercials and around 1999 we got our first feature film project.
After this I spent some time just doing TV spots, a lot of quick turnaround projects as a 3D generalist. Interested in working at a larger studio, I took a job at Guerrilla Games. It was great to learn how problems are solved in a game studio working on a Triple-A title. From here, I made the jump to the US. I started in the effects department of DreamWorks Animation and stayed there for about 5 years. After a beautiful week at SIGGRAPH in Vancouver, I decided to take a job at Image Engine to work on ELYSIUM. After finishing ELYSIUM I started on the sixth instalment of the FAST AND FURIOUS franchise.
How did Image Engine get involved on this show?
Image Engine and Visual Effects Executive Producer Steve Garrad have a long-standing relationship with Universal Studios. When FAST AND FURIOUS 6 started going through the pipe they gave us a call.
How was the collaboration with director Justin Lin?
Chris Harvey our VFX supervisor met with Justin several times in person. Talking to Chris, he really seemed to enjoy the time he got to spend with Justin – every time he brought back a lot of Justin’s energy. I remember one occasion where we sat in our theatre for an entire morning. Chris, Kelvin McIlwain (production VFX Supervisor) and Justin went through every shot in painstaking detail. Afterwards everyone was excited and had a very clear direction on what Justin was looking for and how to take each shot towards final. I am sure Geoff Anderson, our VFX Producer, had a mild heart attack. When you put 3 creatives together to discuss each shot you are pretty much guaranteed new ideas come up and work gets added. In the end the shots were better because of it.
What was his approach to the VFX?
For Justin when it comes to the FAST AND FURIOUS franchise, nothing is too big! He wants more and bigger action. Beyond that he really focuses on the « filmic » quality of the shots, always looking for subtle cues in the final image that invoke the sense that it was all caught in camera and on location.
What have you done on this show?
I worked on FAST AND FURIOUS 6 in the role of CG supervisor. I started by visiting the set in London, making sure we gathered all the information we would need later on to recreate the location. After getting back to Vancouver, I worked with R&D to figure out the tools and pipeline the 3D departments would need to deliver this show.
While in production I was supervising the CG departments, mostly dealing with lighting and effects. The parkade sequence was a big chunk of the work for them. Other sequences in the film required digital set extensions of a prison, creating Tokyo from scratch and making a location shot in London look like it was right next to the Kremlin. This included creating the aftermath of a car being launched into a building – and taking footage from a drone flying over Moscow, then completely rebuilding the shot as a matte painting and adding a helicopter.
How did you approach the parkade sequence?
From the beginning it was clear that the « Shaw Trap Raid » sequence would be the biggest challenge for 3D. This sequence involved a parkade collapsing while Shaw was racing out to avoid the police raid.
We spent a considerable time developing new tools and workflows to make sure we could destroy a full sized building with enough detail and direct-ability. In parallel with building these tools Sam Hancock, the FX Lead, created a rough pass of the full collapse. This was one long simulation of the entire event. We handed the renders of this over to the editors to allow them to cut to the events. From here on we spend all of the time in FX refining this first pass.
In the meantime Mathias Lautour started to figure out how to light and render these scenes. The main challenges here were the big amounts of geometry that would need to be rendered, and the tight integration with FX. Mathias was joined by our lighting lead Caine Dickinson and together they developed the look and setup in the trailer shots.
Image Engine has always had great success with 3Delight. For this show, we stuck with this strategy, but instead of the usual pre-processing and caching of shadow maps and indirect lighting point clouds, we switched to using 3Delight in full raytrace mode. Since this was still under heavy development we ran into several issues. Daniel Dresser from the Image Engine R&D department was crucial in ironing out some of these bugs so we could actually deliver some huge shots using this setup.
What was the real size of the set?
The location they found was in Canary Wharf, London. This was the foundation of a big tower that did not get finished. The whole set was roughly 85 x 85 meters and two floors tall.
Can you tell us about the creation of the environment?
From the set, we brought home a lot of reference photos and a beautiful LIDAR scan of the set! This helped us in determining what we had to build in 3D and where to place it all. While we were on set, the idea came up to replace the shallow pit on the location with more floors below to give the impression of a much larger parkade. This meant every shot with the parkade needed the set extension.
Did you create CG cars for this scene?
Most of the police cars parked on the upper level of the parkade would eventually end up with all the rubble at the bottom. We created digital versions of all the specific models so we could simulate and render them together with the crumbling building. One of these cars, the Astra, had already been built by Double Negative and they shared it with us. We ended up creating about four different models of police cars.
As he escaped, Shaw destroyed the car park. How did you create this massive destruction?
The FX department at Image Engine mostly uses SideFX’ Houdini. Using Houdini allowed us to prototype, test and finalize many of the required tools without having to write a lot of code. One of our developers, Andrew Kaufman had just developed a great system to help us manage all the versions of these tools. This allowed us to keep refining the pipeline even when we were already delivering shots.
As the shots progressed we realized that almost the entire set would need to be replaced with a digital copy. So assets added all the objects one would find on a construction site (the parkade was still under construction). The building would then go to FX where the floors, columns and walls were broken up into interesting shapes.
The lighting and rendering of the dust was handled in effects, inheriting the lighting setup from the rest of the set. FX made sure to provide a lot of extra passes, like separate shadows, split out lighting and several masks that would allow compositing later on to sit the dust into the place and CG renders. The compositing team lead by Jesus Lavin did a great job integrating the lighting and FX renders.
Have you enhanced the on-set explosions?
The practical explosions on set worked out great and most of them are in the film as they were shot. However, later on in the sequence, when the upper floor is collapsing, the destruction needed to integrate with the practical explosions. This was mostly solved with a lot of love from the compositing team and some elements from FX. Since only a few of the columns were built and rigged to explode on set, we needed to add CG explosions. In several shots these CG explosions are right beside the practical ones on screen – the FX team did a great job.
Can you tell us more about the various simulations for the destruction?
The full destruction of a shot would be the sum of several simulations on top of each other. The first step was a rough fracturing and rigid body simulation, these were fast and required a lot of tweaking since they defined the choreography of the destruction in a shot.
Another large visual element is the dust. The motion of the dust was mostly driven by the motion of the initial rigid body simulation. Once we were happy with both of these large elements, effects started to refine them. The bigger pieces would get a secondary fracture and rbd pass, breaking into smaller pieces when they collided, etc. In this pass we added in all the set pieces, the police cars, scaffolding and some simulated cloth. On top of this, we added a layer of small rubble and debris to add the fine detail. All of these passes would also spawn their own dust trails that were added to the main dust.
The car park is full of cops. How did you populate the environment during the destruction?
We replaced all of the policemen on and near breaking parkade areas. The police on set were very much focused on capturing Shaw. They showed little or no reaction to all the destruction around them. When we started to add digi-doubles falling two stories amongst a lot of debris and explosions, it started to feel really weird that their colleagues standing right next to them did not react, so in most of the shots, we ended up replacing all the police. Sebastian Weber, Earl Fast and Jenn Taylor did a great job hand-placing and animating all the police falling, being hit by the cars and reacting to the mayhem around them.
Did the night aspect help you on this sequence?
The night aspect helped in the sense that Justin was looking for really dramatic lighting. Having all the spots and local lighting play an important role in helping us achieve this. Most of the scenes were shot with a large light flooding the set. We found we had to dial that down quite a bit to achieve the dramatic look. Since it was dark we could have fun with the police lights and headlights illuminating the dust as they fell in.
What was the biggest challenge on this project and how did you achieve it?
One of the biggest challenges was the amount of departments, people and technology involved in creating the collapsing parkade. Making sure everyone was on the same page, making different renderers integrate well together and that the massive amount of data flowing between departments is received well turned out to be a challenge.
We mostly solved this by seating everybody together. As for the technical side, it is all solvable, but using multiple renderers in one shot (3Delight for the hard surfaces and mantra for the dust) is still a big pain and making sure it is all lit properly and interacts nicely takes a lot of effort.
Was there a shot or a sequence that prevented you from sleep?
We started working on this project in December, knowing there would be quite some preparation involved. When two of the biggest destruction shots were requested for an early trailer delivery, we needed to shift gears and get going fast early on.
Pau Rocher was compositing these trailer shots and 3D was throwing renders at him as fast as they could get it out. He managed to create some great looking shots out of this chaos. Getting everything sorted out two months sooner than originally planned caused a few sleepless nights, but the upside of this was that for the rest of the show we were in great shape.
What do you keep from this experience?
An appreciation of good communication. We need great tools to manage data and productions, but nothing beats people just talking to each other when they are working together.
How long have you worked on this film?
I went on set on September and finished the project in April. Most of the crew started in December, so Image Engine worked on this project for around 4 months.
How many shots have you done?
We delivered around 200 shots.
What was the size of your team?
For most of the production I would say the crew was around 30 people, with about 15 of those working in the CG departments.
What are the four movies that gave you your passion for cinema?
My memories of what I loved and watched when I was really young are a bit vague. The first movie I remember seeing in the cinema is E.T. Some movies I watched over and over when I was young were along the lines of BEVERLY HILLS COP and FERRIS BUELLER’S DAY OFF. A little later I remember really enjoying several Dutch movies, for example THE NORTHERNERS. As a teenager, one particular movie I remember making a big impression on me was A CLOCKWORK ORANGE. At that time I was fully hooked on film and loved watching films from the Coen brothers, David Lynch and Almodovar. Work-wise, the films that influenced me most are probably TOY STORY, TERMINATOR 2 and JURASSIC PARK.
A big thanks for your time.
// WANT TO KNOW MORE?
– Image Engine: Official website of Image Engine.
© Vincent Frei – The Art of VFX – 2013