Plus, the making of the Deadpool corps, blood and gore in the opening titles, and the rise of Dogpool, in this breakdown of Framestore’s work for Deadpool & Wolverine.
Working with production visual effects supervisor Swen Gillberg and visual effects producer Lisa Marra, Framestore completed many key sequences for Shawn Levy’s Deadpool & Wolverine.
Among them were shots of the character Cassandra Nova (Emma Corrin) moving her hands through Mr. Paradox’s (Matthew Macfadyen) face (a type of shot also completed by other vendors), enhancements for Dogpool, much of the carnage for the Deadpool corps fight and the oner there, and scenes of the final Timerapper machine in action. The studio also delivered significant previs and postvis for the film.
Here, befores & afters asked members of Framestore’s team–Kaya Jabar, senior visualization supervisor, Matt Twyford, visual effects supervisor, and Joao Sita, visual effects supervisor, about some of these big-ticket items in the film.
b&a: For Cassandra passing her hand through Paradox’s head, what did they film for this? What was Framestore’s approach in showing the hands through the surface of the actor? And specifically what were the 2D/3D solutions that made this so interactive and ‘scary’ to watch?
Matt Twyford (visual effects supervisor, Framestore): The plates were shot with Cassandra and Paradox as you see in the movie with only Cassandra’s arm and hand tucked away behind his back. In the final VFX her arm and hand are fully digital which gave us the animation flexibility to adjust angles and performance for best storytelling. We then replicated Paradox with a matching digital double, with full performance, lighting and skin simulation matched to the plate. This allowed us to decide on a shot by shot basis to blend between our VFX and Matthew’s performance seamlessly. The skin effects themselves were a complicated rotation between Maya and Houdini with the major displacements being handled by the animators for maximum interactivity with Cassandra’s fingers.
We then took that animation into Houdini for the fine detail and skin stretch simulations and then back into Maya for modelling sculpting to hit small creative notes. Then back into Houdini for all the hair simulations and then finally into our in-house renderer Freak using the simulation maps from Houdini to control skin thickness, stretch and bloodflow. The compositing team used Nuke to blend into the live action matching all the lens and camera characteristics and perfecting the blends.
I think an absolutely convincing digital double was the main reason it feels so interactive and scary. It means you absolutely believe it is Matthew Macfadyen who is experiencing this and allowing us to keep the effect itself a little more lighthearted and over the top.
b&a: How was Dogpool brought to life? What dog or stuffie was used on set, and what did Framestore do in adding to or creating this performance?
Joao Sita (visual effects supervisor, Framestore): Dogpool is played by Peggy, a British dog that was cast and trained to perform in front of the cameras for the film. She plays in most shots as the real dog with the performance changes mostly done to add more character to her eyes through the goggles and in few instances has been fully replaced by her digital version in order to get the body performance adjusted or to enhance some of her stunts such as coming through the portal, leaping over to Deadpool’s arms and running in the oner shot.
The brief was to add performance to the scenes where she would be reacting to the other characters’ dialog or would deliver a specific emotional beat. We started by playing the sequence and spotting moments in which we could add a performance to the eyes that would bring her into the scene, so, moments such Nicepool coming out of the portal, we greeted this moment in which she looks surprised and excited which combined with Peggy performance in camera with her body trembling made it read rather genuine. Also when Dogpool and Nicepool discuss to whom she belongs or is in love with, she looks away avoiding getting into the discussion.
b&a: For the Deadpool corps sequence at street level, can you talk about what aspects of the large fight required VFX intervention? How were you dealing with in terms of scans of the characters? What about environment extensions?
Matt Twyford: The Deadpool corps sequence mostly used more traditional compositing VFX although we used many techniques through this long sequence of over 200 shots. Crowd duplication was mostly using custom bluescreen elements and split screens with multiple takes. Full sequence performances were shot using dozens of the Deadpool variants and duplicates and we used Nuke’s particle tools to drop them into our shots controlled by the lidar geometry.
The city environment was a full street block set at Pinewood Studios and we extended the city outwards to horizon and upwards to rooftops with an extensive digital build, filling it with a mixture of digital and blue screen crowd and traffic. For the fight scenes there were many custom full and partial digital doubles dropped into the chaos to allow dismemberments, stunt enhancements, sword/gun action and story continuity along with integrating digital and SFX elements. For the ‘Oner’ we actively body tracked nearly 100 separate performances to allow the fully blood digital simulations to interact correctly and the swords to accurately reflect the injuries.
b&a: For the oner with Deadpool and Wolverine, how did viz help here? What kind of stuntvis or actual stunt performance/mocap was filmed for this? How did Framestore approach its build of the characters — what details were tricky to replicate here?
Kaya Jabar (senior visualization supervisor, Framestore): The oner was the most challenging sequence to tackle in visualization. We started at script level working to the idea that the camera will move slowly, parallel to the action, as if observing a comic book panel. Initially we also tested doing smash zooms into the action but these felt too stylised and we settled on an unflinching tracking shot. We workshopped some ideas for our characters, working out really rough beats of action and the overall feel of the piece. Some of these pieces were then incorporated into a beautiful storyboard that was used to balance out the different gags and decide on the overall beats using iconic silhouettes from the comics. Once the flow was decided, we used our internal capture lab team to run through some of the stunt action in order to speed up our animation process. With so many Deadpool variants working in perfect sync we knew that we couldn’t waste time animating everything by hand.
Once again the process was incredibly collaborative, we pitched small moments to the team, adding a balance of gore and comedy to every second. We knew iterating quickly was key to incorporating all the ideas the filmmakers had for the piece. Once everything felt right we created detailed techvis for the camera move and helped audition different motion control equipment that could achieve the shot without being constrained by camera track length, finally settling on the Milo. For this shot we also designed a specific HUD in Maya that could count the number of actors on screen per frame, to help us visualize how many stunts actors were required at any given moment. We also designed the different elements we would need to make the action feel continuous within the constraint of the sheer number of extras required for the final piece against what we had available. We worked closely with stunts to ensure that the different sections of the shot served the different stunt wire rigs required to help our leaps and throws. Once the vis was in a great place we passed the final move over to SFX and ran rehearsals with stunts and moco, this was a brilliant process as everyone embraced the gags in the vis and worked hard to match the ideas we had worked so hard to incorporate.
I also helped on the main shoot days, supporting the moco team and tweaking the camera move to ensure the original vision remained intact. Once the plates were shot they were passed onto our team to start postvis. In postvis we worked on polishing the piece, adding all the effects and tweaking the timing of specific moments for impact. Our main contribution in postvis was the end of the Oner, where our heroes burst through the glass, we designed the speed ramp and worked out the posing, managing the complex reprojections and iterating through a number of heroic poses and timings. Our close collaboration with our VFX team also meant that we could incorporate environment renders and lit animation pieces into our postvis without compromising on our speed. A standout piece of this shot for me is the opening move from Deadpool, the little “singing in the rain” twirl was something that started in the previs and was a true testament to my team and how deeply they understood the tone of the movie.
Matt Twyford: For such a complicated shot the previsualization was crucial to the planning and execution of the shoot. The plates were filmed as four sections of one continuous motion control move. Each of the four sections had extremely complex stunt coordination and timing with around 30 stunties involved in each pass.There were over 80 passes of motion control shot over two weeks through the middle of a British winter and this is where the postvis was invaluable in experimenting and locking selects and timing before committing to the thousands of VFX days of tracking, paint and roto. We worked in parallel with each other for many months, bouncing cameras, retimes, character animations and renders between the teams. This offered clients fast creative iterations while not holding back the full quality VFX build.
b&a: Tell me about the design elements for the destruction of the Timeripper machine — what were the important things to communicate in terms of energy flow, spaghettification, and hitting certain beats while this action occurred?
Matt Twyford: The Timeripper machine and its environment were full practical builds along with the lower power room our heroes find themselves in. We knew there was going to be a lot of destruction so we created full digital assets of the sets and props to give us the maximum flexibility in the visual effects design. This proved invaluable as even though all three of the characters were going through the same traumatic experience their location and costume meant each had custom approaches.
Spaghettification was one of many simulation developments but one that linked in well with the aesthetics and the Marvel universe science logic. We had designed and built the effect for Loki S2 and were well versed in how we could get the maximum impact from this characterful simulation. The animation was used to visually show the tug of war between Cassandra and our Heroes clearly defining the transition between normality and the unravelling universe. To add even more tension and peril both of the rooms were then slowly destroyed with tiles being ripped off walls and ceilings collapsing to create a swirling volume of texture and depth that enhances the set lighting. The ramping up of the destruction simulations and spaghettification works with the edit and sound building to the crescendo of Cassandra’s atomization. The destroyed CG environments and assets were then used for the aftermath scenes dressed in smoke, water and flames.
b&a: For the opening titles work, can you talk about what kind of planning and previs was done here? How did they then film live-action plates? Can you discuss the need to take stunts and gore further, and how you approached this?
Kaya Jabar: The opening sequence was one of the key creative challenges we faced in previs. We knew that stunts and physicality were a large part of the appeal for the cold open so we relied heavily on motion capture for the previs work. Taking the initial storyboard ideas we used our in-house capture team and cast a stunt performer to create vignettes and workshop ideas for the different credits we see in the final piece. We focused on dynamic camera work and comedy here, trying to find ways to play with Deadpool’s performance and coming up with a longlist of ways we could use the skeleton to dispatch TVA agents.
The team also worked at higher frame rates out of Maya, using in-house tools to render overcranked, allowing editorial to create dynamic speed ramps in the edit and tweak the timing of each hit. After our initial pass on the sequence, the stunt team rehearsed the gags and further developed each key moment ensuring everything remained safe and achievable. We also created the cornerstone dynamic spinning shot, where Deadpool dispatches a circle of TVA agents, ramping into slow motion for the final hits, adding the joke at the end of the agent exiting the time door and getting smacked in the face. Data and techvis for this move was then provided to SFX who created a circular speed rig allowing the camera to move at the pace of the previs. The process was incredibly rewarding, we were tasked with coming up with more and more outrageous ideas, everything was approached with a “Yes and” attitude allowing us to create the ultimate Deadpool spectacle.
