VFX supervisors Xavier Bernasconi and Simon Maddison talk muzzle flashes, face replacements, and a cacophony of van, motorcycle and helicopter crashes on the Amazon MGM Studios action-adventure thriller, starring Jason Momoa and Dave Bautista, now streaming.
What do you get when you cast Jason Momoa and Dave Bautista as estranged brothers trying to solve their father’s murder? An action-adventure thriller appropriately called The Wrecking Crew. The film, directed by Angel Manuel Soto on behalf of Amazon MGM Studios, now streaming on Prime Video, features a high-speed Honolulu highway chase that involves a Chrysler Pacifica pursued by motorbikes and a Eurocopter helicopter. There’s plenty of gunfire, destroyed cars, and an explosive crash. The signature sequence was among the 354 visual effects shots delivered by DNEG, who was also given the task of doing digital doubles and face replacements.
Enjoy the VFX breakdown reel, then learn more about how some of the work was produced:
While Production Visual Effects Supervisor Stan Alley worked with the main unit in New Zealand, Xavier Bernasconi, DNEG VFX Supervisor, looked after the second unit and stunts; Scott Anderson joined later as an additional production Visual Effects Supervisor in post to get the project across the finish line. DNEG’s contributions were divided between facilities in Sydney and Vancouver, with the former being led by Simon Maddison, who assisted with the face replacements, and the latter looking after the Honolulu highway chase under the direction of Xavier Bernasconi. “A lot of the face replacements were done through the new DNEG face replacement pipeline using sister company Brahma’s technology,” explains Bernasconi, Visual Effects Supervisor, DNEG. “We already had the rigs and digital doubles for Dave Bautista and Jason Momoa for the chase sequence, but it was supposed to be inside the car, so the rigs were not extremely detailed. However, when Will Reece [Production Visual Effects Producer] asked us to takeover this work we jumped in quickly, did a lot of work on the rig, and while the face replacement team was doing the plates that we would then composite in, the body tracking and lighting teams did a tight body track at shoulder and neck level. In those types of shots, Scott Anderson was more interested about the flow of the head being natural than focusing on the facial expression necessarily. We found that if the flow was there and you felt it was natural then because of the extreme speed of the shot you wouldn’t focus on the expression; that would come after.”
“Once the shots requiring face replacement were called out, it really was just a case of ‘make them look real,’” explains Maddison, Visual Effects Supervisor, DNEG. “Because the stunt doubles were such a close match to the actors in terms of skin tones, we did have a lot to work with as far as that went; if the skin, texture, lighting, and sweat matched perfectly to the performer shot in-camera, that part was done.” It was critical that the performances felt completely authentic. Matching the actors’ likenesses was only part of the challenge, reproducing their facial performance with absolute fidelity was equally important. Maddison, added, “Rather than simply inheriting the stunt performers’ expressions, we combed through extensive B-roll of the principal actors, building a curated performance library. For each shot, we selected material that best reflected the emotional intent and physical nuance required. When a shot reached animation, it included not only the tracked camera and object/face solve, but also a defined performance target. This approach was time-intensive, but it ensured the final result genuinely belonged to the actors rather than an interpretation layered over a stunt double.”
The film’s compressed production schedule was the primary challenge. “It was important that the director never felt constrained to a single creative option on any given shot,” states Maddison. “To achieve that, I structured the pipeline so that multiple departments could run in parallel wherever possible. The first step was building a curated B-roll performance library from the actors. For each shot, we selected and presented a specific performance to the director for approval before animation began. Once approved, animation commenced immediately, even before camera and body tracking were complete, using a rendered picture-in-picture approach to validate the digital performance. This meant animation was effectively running ahead of the broader shot pipeline. By the time tracking, prep, and often lighting were completed, we could slot in a performance that had already been creatively approved. As a result, there were very few instances where animation notes surfaced at the first comped presentation, as the performance itself had already been signed off upstream.” Some amazing tools were utilized. “The system used machine learning to give us not just a 2D replacement but a full 3D animated head,” Maddison explains. “Because of the way the tools worked, we also retained full control further down the track. And they were all built on the real actors' expressions, meaning the performance we got was that of the actors, not an interpretation.”
The Honolulu highway chase combined stage work with the principal talent and live-action stunt photography captured in New Zealand. “The complexity of that sequence and the fun challenge of it was to work out where at any point Jason and David would be in the sequence geographically,” notes Bernasconi. “The first thing that we did was to create a Gaussian splat representation of the highway from drone footage. We used that to reverse engineer, starting from the end of the sequence all the way to beginning, and based on an average speed for a van, [to figure out] where they would be. And let’s not forget the story! Editor Michael McCusker had a hard task making it clear what was going on and not just a cacophony of crashes, and why they were chasing them. Our director was clear on what he wanted and where certain beats had to be. We had those as our anchor points, such as when the Yakuza gets thrown out of the van and when the motorcycle has a head-on collision. Everything had to be slotted in between.”
Helping provide a sense of continuity was the postviz, with some of the camerawork being altered later on. “We did a lot of work in our layout department in terms of continuity for the cars and crashes,” Bernasconi says. “The van goes through the crashes that are caused by the helicopter way before the motorcycle goes through. We had to make sure that those reference points were still there or by using natural elements like the rock formation or car colors or types of cars.”
Plates were used as reference, but then the camerawork, timing and damage progression of the van were changed. “At the end we decided that it was easier to redo it in CG than to shoehorn a particular plate,” remarks Bernasconi. “We shot the van on a dragstrip, which was helpful as reference. There were some crashes that were kept as a plate because they were cool, but we added cars on top of it. All of the exteriors are CG.” Six miles worth of CG environment had to be created with vegetation typical of O’ahu Island. “The tricky part is vegetation has this ability of looking fake if the set dressing and scattering is not well-thought out,” Bernasconi adds. “Plants change throughout altitude. You start to go into lower-height trees when you get higher up. There are more bushes and rocks, including more palm style vegetation situated near the highway. Scale is important. On top of that is the movement of the foliage, which was added and integration with the helicopter because when it goes whooshing by, we had to have motion on the trees as a byproduct of the downwash caused by the propellers.” DNEG also had to content with a big switch in the lighting that had to be accommodated within a 10-week period. “Up to June 2025, the lighting was supposed to be moody like misty,” Bernasconi says. “Then there was a change to have pockets of sun, light, and shadows to give the idea of speed and motion, as well as the idea of going from a more claustrophobic beginning where they feel trapped to finally getting up towards the safety of the tunnel, which is lighter. It was tough.”
Muzzle flashes are a secret artform. “We do them in CG not 2D and have body tracks of every single action and plate in order to do an interactive pass in the lighting of the muzzle flash on the body or car,” states Bernasconi. “We can then overlay that additional lighting pass on the plate. I’m a big believer in eye reflections and interaction, so we always track spheres on the eyes and roto eyelids. If there is a blink, we can occlude the reflection, render reflection for the eyes because with bluescreen, the reflection will not be there for them travelling through and trust me, you do see changes in the eyes if you’re travelling. There is modulation on the highlights and reflections. All of these small details make or break an effective muzzle flash. And of course, every gun has a different muzzle flash.”
On a much larger scale is the explosion and destruction caused by a crashing helicopter. “I have been in a helicopter a couple of times for a shoot or transport,” Bernasconi says. “They are very wobbly machines; their behavior in the air is difficult to nail because their momentum, acceleration, and rotation is slightly off. And the same applies when the helicopter crashes. We were trying to combine the creative vision of our director with believability.” One challenge is conveying the proper sense of weight. “When it lands on its side, the propellers chop into the concrete, so we have concrete breakup flying everywhere, and damage,” Bernasconi continues. “We tried to portray the consequences of such a huge machine landing on a concrete pad. Concrete is hard but the spectacle needed to be portrayed.”
“We had a bunch of difficult shots for one reason or another,” he notes. “However, the complexity of this particular work was the length, layout and continuity of the sequence. Creating a continuous believable action packed three and half minute piece that made sense. Thankfully, xSTUDIO Timeline [DNEG’s open-source review tool] was a godsend because I was able to consistently load the entire cut with updates from layout and lighting. Having different tracks for layout or lighting or compositing or animation helped to ensure that the cut made sense because the continuity was there in a responsive way. I was able to send that as QuickTimes to our client because in that way we were able to shorten the review process. We sent so many versions. We would send the entire sequence almost three times a week re-rendered. That was a lot of work, but it was the only way to help Scott and our director to keep the pulse of the sequence.”
Bernasconi concludes that ultimately, some of the final scene work contributed to the narrative. “We had to come up with all of the actions of the surrounding cars and how they crashed because they weren’t scripted or designed; that was the fun part and really creative.”
Trevor Hogg is a freelance video editor and writer best known for composing in-depth filmmaker and movie profiles for VFX Voice, Animation Magazine, and British Cinematographer.
