The Academy of Motion Picture Arts and Sciences announced that 20 films were in the running in the Visual Effects category for the 92nd Academy Awards. Most of the films we have covered and some newer films we will be posting this week. Additionally, for major films we covered a while ago, we’ll be posting special fxpodcasts with the supervisors in the coming weeks. The list, in alphabetical order, is Ad Astra, The Aeronauts, Aladdin, Alita: Battle Angel, Avengers: Endgame, Captain Marvel, Cats, Dumbo, Fast and Furious Presents: Hobbs & Shaw, Ford v Ferrari, Gemini Man, The Irishman, Jumanji: The Next Level, The Lion King, Men in Black: International, Midway, 1917, Spider-Man: Far from Home, Star Wars: The Rise of Skywalker and Terminator: Dark Fate.
This has now been narrowed to 10 films for the bake-off, (in alphabetical order by title), are:
- “Alita: Battle Angel”
- “Avengers: Endgame”
- “Captain Marvel”
- “Gemini Man”
- “The Irishman”
- “The Lion King”
- “Star Wars: The Rise of Skywalker”
- “Terminator: Dark Fate”
It is the last of these we explore below with Digital Domain’s Jay Barton. ILM was a lead facility on Terminator: Dark Fate and Digital Domain worked on three sequences on the film comprising of a total of 176 shots over approximately 5 months.
- The factory fight between Grace and the Rev-9,
- The Helicopter Attack sequence at the derelict industrial complex sequence
- The C5 Galaxy airborne fight between Grace, Arnold and the Rev-9 (the weightless interior fight)
The work required “some complex hero head replacements for all of the main characters at one point or another,” explains DD’s Visual Effects supervisor Jay Barton. “As well as full-body replacements for all of the main characters in different sequences, – with slightly different methodologies depending on what the performance was or whether we were trying to match a digital head to a stunt performer or if it was a fully synthetic digital actor”.
ILM provided the transition effects and development assets for the main characters.
While ILM did much of the hero cast scanning and provided the data to DD, the factory fight included a reasonable amount of shots of Dani Ramos’s father (Tristán Ulloa), which are when Gabriel is impersonating him.
For this sequence, Ulloa was on set, so DD did not require a detailed scan and model of Ulloa’s face. “When we had him being shot in the head with a shotgun, we had his performance. We were able to keep and only replace chunks of his face as necessary”, commented Barton. The team still needed to align the skull base to fit the actor’s facial features and track the liquid metal effect. “But the liquid metal that was on his face was more a kind of putty or clay than an actual liquid,” he adds.
To add the liquid metal effect the team did use camera projections to help with the alignment and transition, and having a good estimate of the actor’s face really helped with the object tracking. “The thing a lot of people don’t realize is just how good of a model you need to be able to track properly for visual effects on faces. You really have to have a good model for your own sanity to make sure that what’s happening is accurate. You can’t just take a generic head shape roughly line up the eyes and the nose – it just doesn’t work”, explains Barton.
For the fight at the industrial factory location DD needed to deliver a series of digital double shots of Gabriel jumping or flying through the air or “doing action sequences that no one wanted to do with wirework” Barton adds. There are also quite a few of the fight sequences where Gabriel is armed with his double blades. These shots required body and arm tracking and in some cases Gabriel’s character was a stunt performer with additional face replacements or full head replacements.
The location had been LIDAR scanned by the production. “We got a full package of LIDAR for every sequence, every location, all the sets, and actual locations,” Barton comments, adding that, “we then used our own in-house tracking software to generate the geometry that we needed: object and camera tracking. From ILM base, the DD team got assets with lookdev for our layout of the factory set and these were then modified and adjusted for DD’s lighting pipeline”. The explosion Houdini FX team was led by Eddie Smith.
ILM did the majority of the work on the exterior collisions and crashing into of the hull of the C5 plane.
While there was some wire work, the very nature of the plane’s motion worked against shooting primarily wire or rig work. The sequence was complex as the fight takes place in a changing landscape. “Sometimes the plane was banking to one side and they’d slide one way, or they’d slide forward and then go into zero G. None of which lends itself to wirework, so we had to use digital doubles and other complex approaches,” Barton explains.
The other complexity is that shots often transitioned from one plate to another, so the DD team had to seamlessly connect the elements and unify the environment. The interior of the plane’s cargo bay was about 70 feet long. The production built one section and then they would re-dress it depending on if the camera was meant to be looking aft or forward, adjusting the lighting to match. The production even hung a blue screen so that the plate could double as a shot in the middle of the cargo bay, with debris flying and rushing air from the descending plane.
One of the most complex shots started in the cockpit and seamlessly moves from Grace (Mackenzie Davis) as live-action to digital double, to CG environment. Shortly after the collision, Grace has to put the C5 on autopilot and run back because she’s heard Danny scream out for help. “It is one of those changing gravity shots where we, the audience, are with her in the cockpit, looking at the actress. She gets up, turns and the camera pans with her as she goes towards the back of the plane,” outlines Barton. The set is very small and it did not extend back to the place where a stairwell would have gone down into the cargo hold, so DD took over and as the shot follows her it becomes digital. “It goes right behind her shoulder,” he adds. “She goes down the stairs and the camera pauses for a second behind her head to survey the damage, and then the camera goes between her arm and her head to come right around the front of her face, and we see it resolves into a closeup shot of her face as we register that she has seen what she has to do”. Finally, still as part of the same shot, she lets go of the stairs and the camera pans, following her as she drops down into the center of the cargo hold. All as one complex (multi-setup) shot.
To achieve this shot, in addition, to some blue screen work, (window replacements, etc), the production shot on a loose spring stage with Davis filmed by four high-speed cameras, all framed on her face. “Then we got a bunch of different takes of performances and worked with the director (Tim Miller) to figure out exactly which part of the performance he liked and what would sell the story,” Barton explains. DD then recreated Davis’s facial performance and projected bits of the live-action footage onto the CG model blend between their digital double and Davis’s live-action performance.
DD decided that while their digital double work was looking good, mixing in live-action partial faces would sell the shots. “We used bits of Mackenzie’s face that are transitioned in and out over the full CG,” he recalls. This was mostly done for realism but also if the director needed the team to alter a performance. “We had to keep the eyes as all CGI, because of the reflection of everything around Grace, but the rest of her face is transitioning in and out of the four camera views, to match the lighting of the environment”.
ILM did the primary facial scanning and provided exports of ILM’s base shapes, along with some of the original Medusa scans, but DD had to make their own face shapes to work with for their pipeline. The skin texture was built from the unwrapped skin photograph files they received from ILM. Grace’s eyeballs were DD’s own in-house models made to match Davis’s eyes. DD did do their own simple on set stereo photography of the actors, and texture reference photography.
Digital Domain rendered in V-Ray and some Houdini sim work was rendered in Mantra. The material was mastered out at 2K resolution.