What have I become
My sweetest friend
Everyone I know goes away
In the end
And you could have it all
My empire of dirt
I will let you down
I will make you hurt
– Johnny Cash
Hugh Jackman decided that he wanted Wolverine 3 to be a mix of Unforgiven (1992), The Wrestler (2008) and Shane (1953). Director James Mangold provided the stage, for him to deliver this moving performance.
Production VFX Supervisor Chas Jarrett supervised the 1100 visual effects shots from Image Engine, Rising Sun, Soho VFX, Lola, and an in-house team. The film was pre-vized by Halon, headed by Clint Reagan.
Digital Logan: beyond head replication
Image Engine handled most of the digital doubles for the film, doing the primary work on both Hugh Jackman and Dafne Keen who plays Laura. The company completed 299 shots, mainly full digital head replacements.
Much of the work Image Engine did in the film is quite subtle, according to Martyn Culpitt, Image Engine’s VFX Supervisor on the film. “It is not like there are dinosaurs running through the forest,” he comments. While the film is graphic and clear, for safety reasons it involved visual effects. Much of the most complex work Image Engine did is invisible head, face or body replacements to allow for the film’s complex stunts and hard core action.
The Canadian team completed 63 head replacements in 51 shots, as some scenes involved multiple characters in the same shot. The company also did ten full digital body doubles and then environments, explosions, and set extensions.
“There was a lot of complex stunt work done…full contact work, when we were on set the stunt team were fully hitting each other,” explains Culpitt. In addition to this, in many of these scenes there was a lot of craw replacement/adding and of course blood etc from the fights. “We did a lot of CG blood, CG gore and, of course, enhancing wounds (and practical makeup),” explains Culpitt.
There were four stunt actors, with two primary stuntmen Eddie Davenport and Daniel Stevens playing Wolverine/Logan/X24. While one might think that since every stunt double is being replaced with the Logan character, there is only one Logan head. But in reality that was not the case. “We had to build a lot of versions, as the proportions have to match the stunt actor,” Culpitt explains.
Most readers may now be familiar with the process of scanning an actor like Hugh Jackman. Most high end digital humans use a similar path, which is along the lines of a USC ICT Lightstage scan to get the skin reflectance properties, a FACS expression rig is built, and then this is hand animated or motion capture driven. While this is impressive technology, fxguide has covered this part of the pipeline extensively in many recent stories.
What is not so well known is how that digital head replacement works in terms of the stunt doubles. It is normally skipped over how the team combines this digital double with the live action photography and yet this part of the process is vital to selling a digital head replacement of Logan.
Detail Stunt Double Replacement: not really Hugh’s head
“The whole issue of using stunt doubles was a really big deal from the very start of the film,” recalls Jarrett. The process of replacing a stunt double in a film of this quality pivots on first getting an accurate match to the actual stunt person. Each stuntman and, of course, Hugh Jackman, were scanned and digital assets built.
Take, for example, a shot where Logan is running through the forest trees to kill attacking bad guys or “reavers” at the end of the film. This sequence was nicknamed ‘Logan’s Run’ by the production. What seems like one long shot of Logan running and impaling a bad guy on a tree is actually four separate takes of different stunt guys running, combined with head and neck replacements, digital claws, background fix up and CG gore.
While there is enormous skill in getting a turntable shot of Hugh Jackman looking real, for Image Engine it is just as important to focus on the faces and bodies of the actual stunt doubles.
In Logan’s Run, the shot starts with Hugh Jackman and then transitions to stuntman Eddie Davenport for the next two takes, where the stuntman’s heads and claws were digitally replaced. The final fourth part involved an entire digital double, removing any actors, and adding a fully CG Logan jumping and killing the bad guy.
The running sequence was shot from a camera motor bike driving through the forest as Jackman or Davenport ran. “It is really complex as Logan needs to slash and kill someone – but he cant do that properly half way through a step, he needs to complete the step and then do the kill,” explained Jarrett. On-set, Logan was also flown across screen using a complex wire rig “but it looked like he was pivoting on his hips so they wanted that replaced,” explains Culpitt.
Why are we modelling the faces that are being never seen?
There are two reasons to accurately model the faces, bodies, and, in particular, the necks of the stuntman doubling for Hugh Jackman. “We build four different Logans; each has to match the proportions of the stunt actor. We did a lot of work to get that right”, explained Culpitt. If the Image Engine team simply replaced the face area of the stuntman, in what is sometimes called the hockey mask solution, the shot would almost certainly fail. Since the film Titanic, companies have tried to sell face replacements by using the hockey mask approach, but they often look unbelievable. The problem is that it is very hard to articulate exactly why they fail.
Image Engine believes the hockey mask approach fails because the head, and ears are highly identifiable as Hugh Jackman. While we, the audience, look to the actors face and their eyes in particular, a part of our brains links a face with the skull or silhouette of the actor’s head. If you know the person, in this case Hugh Jackman, then you will have a cognitive dissonance between the right face but wrong shaped head. But it goes further.
The ears on one’s head are as unique as your fingerprints. Getting Hugh’s ear wrong may not be something an audience member could point to as being ‘fake’ but it would reduce the believability of the character over all. Finally, even if the head, ear,s and face are a perfect match, the proportion of the head to the body and in particular the length of the neck is critical and this is why Image Engine needed a unique solution for each stunt performer.
The team would often get the shot tracked with Lidar so that the camera position was perfect on the wildly moving action sequences such as running through the forest. To make the action dramatic, multiple takes were combined so this step along was complex. Some passing trees close to camera helped wipe the environments, but Image Engine still needed to do a lot of work to make the background appear as one continuous shot.
Once a continuous background plate was made and cleaned up, the stunt performer was object tracked. To help with this, on location, Image Engine placed markers on the stuntman and the stunt women who were doubling for Logan and Laura. This is really part science part rotomation as the animators using the stuntman specific scan and model, try to overlay the most accurate CG model on the action of the scene.
This is made even more complex as most scans are of the stuntman in a neutral pose. But when a stuntman of Hugh Jackman is acting and running, their muscles flex in ways they did not in a relaxed neutral pose. Ad, of course, they are breathing heavily, with all that involves.
“We did a lot of renders around the head only, or just the cloths only, or wire frame to make sure everything was tracking correctlym,” explained CG Supervisor Yuta Shimizu. His team had four full Logan rigs alone to keep track of use depending on the stuntman who was performing. There were additional rigs for each stunt woman who played Laura.
The scanning was done on set by Capture Dimensions, based in Dallas, Atlanta, LA and Albuquerque, do mobile or studio scans. Those mobile scans were “to the skin” to get all the anatomical dimensions correct, comments Culpitt. These scans had blendshape muscles to get correct muscle activation flexing for when they turned or ran.
Test render of the stuntman
Once the team have a perfect line up, they make a digital head and neck of the stunt person and render that into the shot. This is a technical exercise of the the head of the stuntman on his own body, The team has a few issues to solve here for lighting. On location the team walked a mirror + grey ball down the path that the actor ran but there was some bracketing and any HDR for Image Based Lighting(IBL) does not provide the continuous moving lighting reference the team needed as the camera flies down the forest. The LIDAR scan also informed the lighting.
In a new approach, Image Engine also got high speed Phantom footage in the light stage at USC ICT along with a 3D rig of the actual Light Stage and where each light is in 3D space. They did this so they would have one frame for each of the single lights in the Light Stage. With this pass of individual lights they can see the way dappled light spotting through the trees might affect a head.
For example, when Logan gets a momentary spot of light on the back of the ear, the team has reference of what one isolated light would look like, just hitting behind Logan’s ear. To get the lighting on the CG heads “we had to use a lot of Gobos used on our (CG) lights in that particular scene, to break up the light,” comments Shimizu.
Having solved the lighting, the team can still not just place the correctly lit head of digital Hugh Jackman on the body of the stuntman, as the proportions would be wrong. “Even if we put Logan’s head on the stunt actors body and match the stunt actors head,” explained Culpitt, “it wouldn’t be correct, even though the proportions of the stunt actor are correct (for him) it didn’t feel like Hugh. So we had to spend a fair bit of time getting this solved.”
Dont track the face – at all !
“We learned that facial tracking markers are much much less important than we thought they were,” comments Jarrett. The problem with proportions is the reason that addition to putting markers on the stuntmen’s faces, Image Engine moved to putting markers on their shoulders and torsos.
“What we learnt early on is if we tacked the face we would get a really good track, but the face replacement wouldn’t work, since the shoulders weren’t locking,” Culpitt said. “If you locked the shoulders and the head separately you would get this weird stretching of the neck. So we threw away tracking the head and tracked the body. We track the shoulders then the neck and just then put our head on, cause while our animation might reference some of the rotation and the position in space of the Stunt double, but most of the time it is a fully animated head.”
This has one other advantage. All the motion of the head, and thus the motion blur, is being driven from the body, just as it would be if such a digital Logan existed on set. It is more important to drive the head correctly from the body, than to just match to the original stunt doubles motion blur. “It was really tricky to track all that – it took a lot of time,” Culpitt related.
The problem is that relationship between head size, neck and body. “The way Hugh’s head sits on his body, his neck is a bit shorter than the stunt guys, and the width of his shoulders…it just didn’t feel right,” saysCulpitt. “We went around numerous versions with Chas (Jarrett) the overall visual effects supervisor.”
To solve the problem, they discovered they needed to modify Logan’s head to match the body. In a sense Hugh Jackman’s head works on Hugh’s body, but each stunt actor needed their own Logan head, with customised proportions, neck length and features, for those stunt doubles to be read on screen as Logan and identical to Hugh Jackman. “He also has quite a thick neck and strong traps (back of the neck trapezius muscles),”Culpitt adds.
To achieve the level of realism Chas Jarrett wanted, “everyone in the past has sort of done hockey masks, about 90% of our work was full head (and neck) replacements.” explains Culpitt.
There was finally work to be done to make sure all the different stuntman from the 4 different plates of Logan’s Run had matching wardrobe and blood/scars on each stuntman.
“That shot started comping it on set – to make sure we had everything and went right to the end, 9 months later” explains Jarrett. “They finished pretty much when we finished the film… yeah it was a very big very complicated shot.”
The face work on this film built on face work the Image Engine team had built for the film Point Break, but they rebuilt their pipeline for Logan during a large 3 month FACS and rig face building pre-production phase. Interestingly the FACS expressions were not initially build on Hugh Jackman’s face. The first FACS expression session was done at Pixel Light effects, with an animator from Image Engine.
This ‘test run’ allowed Image Engine to build a complex new FACS rigging pipeline before they had the studio FACS session with Hugh Jackman and Dafne Keen at USC ICT in LA. The new rig pipeline is primarily a blendshape rig, but with three layers of controls for the animators. The faces were lit in Image Engine’s Gaffer with both OpenOSL and Arnold shaders all rendered in Autodesk’s Solid Angle Arnold.
Laura is played by young actress Dafne Keen. Laura’s stunt doubles were Rissa Kilar and Marissa Labog (4′ 11″ (1.50m) ). Interestingly, Kilar got two credits on the film for key Laura work and also in the film playing the escaping mutant, ‘Pine Cone Girl’. Kilar was on the stunt team when they did some stunt-viz blocking work for director James Mangold. When the stunt team are suggesting ideas, they just use whoever they have to play the roles that will later be done by the actors. But when the director saw how committed Kilar was, “Jim just said we have to get her in the show” recalls Jarrett and Kilar was also cast as Pine Cone Girl in addition to being Laura’s double.
Cheramie Martin also body doubled for Laura in the Casino sequence where casino rules did not allow the 11 year old Keen to walk. Martin is over 18 but of a similar build to Keen.
“Even with all the correct permits and everything, she was not allowed on the Casino floor,” says Jarrett. This meant that a combination of Cheramie Martin and green screen were needed to have Laura in the casino. The seamless compositing was handled by an in-house team of just three artists who did about 300 shots directly for Jarrett.
“So all the shots of Laura in the Casino were shot green screen… and a few shots from behind her in the casino are actually Cheramie,” relates Jarret. For example, when Laura stops while walking through the Casino floor to admire some cloths, Keen and the glass plate were shot on green screen and then composited over the plate of the casino.
Jarrett was incredibly impressed with Keen. “She is 11 going on 25,” he relates. “She just has such a great world view, and it was obvious from her first audition tape sent over from Spain. Her parents shot her “just prowling around her living room for about 15 mins,” says Jarrett “It was really still and she didn’t say anything, but she just sort of stood there or climbed on shelves and it is just completely instantly obvious that she was perfect.”
Pulses and Claws
Rising Sun Pictures (RSP) worked on one of the more visually arresting sequences which involved Professor X’s application of a “psionic blast”. Initially this was conceived as a huge pulse of energy, but director
James Mangold and vfx supervisor Chas Jarrett evolved the idea from the initial pulse wave idea to a mind control field that caused people to become immobile. Unable to move and breathe or function properly, the psionic blast would eventually kill those around him.
Day one of filming was the Vegas hotel sequence when Wolverine fights against the psionic blast and makes his way up the hallway. The evolution of the visual effect can be seen in the way Hugh Jackman leans into the effect (using a hidden wire rig) and uses his claws on the wall. The original script said that there were waves of energy coming off Professor Xavier, explains Jarrett, “much more akin to the kind of superhero powers you might see in other films. But Jim (Dir. James Mangold) and I talked about not going down those sorts of paths on this film, and keeping everything much more naturalistic.”
On day one of filming Jarrett recalls that “everyone was keen to get going and get the claws out” and the effect was not fully articulated. Thus, in this scene, there was still a sense in which Jackman would need to ‘swim against the current’, hence the claw-wall action.
But Jarrett had already done tests himself, shooting very shaky footage and then stabilizing and felt that the effect could be refined in its logic. There would be motion blur in many of the frames and when the image was no longer shaking, this motion blur takes on a smearing effect that is both organic and very unusual. The team shot the sequences slightly wider than was needed so that shots could be blown up to hide the edges of the stabilizing effect.
“The blast was the most challenging effect creatively,” says RSP VFX Supervisor Dennis Jones, who handled the effects work on the shots. “The sequence was shot natively with camera shake and so there were no clean takes,” he relates. “Initially, we explored effects and treatments designed to add tunnel vision vignetting, as well as applying blur and over exposure, but weren’t satisfied with the results.”
“We found that shots with high contrast content and aggressive high frequency shake produced ideal results without too much modification,” says co-VFX Supervisor Anthony Smith. “We developed techniques to augment the blur artifacts with custom-animated kernels applied through the FFT (fast Fourier Transform) method. That produced sharp and controllable results.”
While the whole frame was given a similar treatment for Logan and the shots of the Reavers attacking Professor X, the effect was modified for shots showing the Professor himself. “As it was coming from Patrick Stewart we changed it to radiate from him, so he is sharp and it spreads out radially from him,” explained Jarrett.
The camera was slung on an Easyrig, which is a back pack with a metal arm that allows the camera to hang on a line at chest height in front of the operator. The production used the Alexa Mini and stripped off as much as they could to make it as light weight as possible. This camera was then vibrated and shaken easily by Jarrett himself standing next to the camera operator.
In the film Saving Private Ryan, the team on that film shook their camera but introduced a reduced shutter to make the action seem very gritty. The reduced shutter removed the motion blur on the shots. Jarrett did consider going the other way and opening the electronic shutter up to 360 degrees, which would double the blur, but a key aspect of the approach is that naturally on some shots the footage would momentarily look normal, so they stayed with a 180 (172) degree shutter.
“What I discovered in the tests was that you need little moments of sharpness, if you don’t get moments of sharpness you end up with an image where you don’t know where to look.” explains Jarrett. “We wanted real sharpness especially on (Hugh’s) eyes about a quarter or a third of the time.” There was a large 20K light outside the window to flood the Vegas hotel set with light and this, combined with a wide angle lens, gave a good exposure. But still it came down to Trevor Loomis, as focus puller, to hold focus on Jackman while they moved and shook the camera. “Trevor – the best focus puller in the world bar none, did it – he is a bonafide genius – I have never come across someone who can pull focus like he can,” Jarrett exclaims.
Interestingly, this was a ‘baked in effect’; the team did not film everything both ways. The team committed and filmed shake on-set into every shot that needed it. Committing to an effect on set is a bold move in a world where people do anything to delay committing to something, but director James Mangold trusted Jarrett and was happy to commit on day one of filming.
RSP also worked on the film doing a lot of claw and gore work throughout the film as well as handled the opening fight sequence.
In the fight, Logan awakes to find his limo being jacked near U.S.-Mexico border. This is our introduction to our beaten fallen hero. Although his skills are rusty and his adamantium claws don’t work quite the way they should, he dispatches his assailants in a swift, violent flurry, stabbing one through the arm and skewering another straight through his head. Visual effects elements included not only the animated CG claws, but also blood, gore, wounds and body part replacements.
Although RSP had created a number of claw effects for The Wolverine, the speed and complexity of the fight proved challenging. “The story and performance beats were well articulated through bash comps from Chas (Jarrett )and editorial mock-ups,” notes RSP VFX Supervisor Dennis Jones. “Still, there was a lot to work out about the mechanical functioning of the claws and how they related to the actors and individual body parts, and we were given considerable license to solve the action in creative ways.”
In addition to RSP who worked on claws and wounds (above), Lola VFX in LA helped with some scar and ‘slash’ cleanup and some de-aging on X24.
SohoVFX handled the desert escape in the limo, completing the robot hand and the CG bikes, and the CG train in that sequence. “I had not worked with them before but are rapidly become one of my favorite vendors,” Jordan comments.
The desert escape was a turning point for Jarrett’s relationship with the director. Logan drives his limo into the compound fence, and, unlike in countless movies before it, he does not knock it over and immediate escape. “It was one of my most favorite moments with Jim in pre-production.” says Jarrett. “I had not worked with him before and at the moment I suddenly got him for the first time. At one stage we did have the fence just being knocked down and he just put a stop on it. ‘No no they cant get through, everyone will expect that – it would be a A-team moment,” Jarrett says, referring to the 70s TV show of the same name. “Everyone thinks that they will just knock through it – but they cant, they just can’t… it was great – and it summed him up on how he handles stuff like that.”