Cinesite, Method LA and Cantina Creative populated Iron Man Tony Stark’s new world with photoreal environments and bold visual effects.

 Part 2 seeIron Man 3 part 1

IRON MAN 3: Brave New World

Method LA was called on to work on ‘Iron Man 3’ quite late into production, starting work in January 2013 and delivering by mid-March. It wasn’t much time for their initial commission of 125 shots, of which just over 80 can be seen in the final film. VFX supervisor Matt Dessero and the team therefore planned to work very accurately to avoid losing time on errors.

Iron Patriot

They were working with assets from three different vendors – Framestore, Weta and Digital Domain, who supplied models of the Iron Patriot suit. Because their shots included those of the suit opening, Method needed to build extra internal parts for the suit, create more detailed texturing and break up the model.


Matt said, “The Iron Patriot model has a lot of moving parts for both the exterior and interior. Rigging these pieces required a bit of back and forth between the modelling, rigging and animation departments, and so we only built what was necessary for our shots. Animation began with a camera track and match move of the actor’s performance. We roughed in the timing with very few simple exterior shapes for the helmet, chest, arms and legs. Getting the opening to flow and feel purposeful was important, and so each animated event was motivated by this performance.”

Making sure the Iron Patriot felt heavy was production VFX supervisor Chris Townsend’s primary concern. He never wanted the suit to feel flimsy. This weight was achieved mainly through the animation, such as adding reverberation as each large piece opened and stiffening the joints to limit arm and leg sway. Once the primary animation was approved, they scored the suit further, and added more movable geometry to the interior and exterior to increase the complexity of the suit opening effect.


Water Tower

A particular challenging sequence involved the villain Savin’s destruction of a water tower through the hot, burning power of Extremis, allowing him to melt down the solid metal of the tower. As he super-heats one of the tower’s legs, their task was to build wide establishing shots showing its instability and ultimate collapse, particularly the  full CG hero shot of the final stages as the tank hits ground and destroys a construction trailer and water rushes toward and flows under the camera.

LIDAR data was supplied of the scene but the challenge arose from the large number of simulations, including the flowing water and rigid body dynamics, plus the amount of animation to complete. Chris Townsend and the team shot a full scale dump tank for most shots of the tower’s collapse. On this sequence and the others on the project, they were using a Maya-V-Ray pipeline with Houdini for effects.


“We also had Chris Townsend’s post-vis to start with, and ran a few physically accurate simulations to understand the timing of the shot. This gave the animators a true sense of speed for the falling tank,” explained Matt. “We quickly realized the amount of time we had would never allow us to achieve the level of destruction that I was looking for from our original full rigid body dynamics simulations for the tank portion of the waterpower. After a couple of weeks into our simulations we received a new edit from production, which was much tighter with less overlap of action across shots. The tighter action allowed us to save time by modelling more blend shapes instead of using lengthy simulations.

Blend Shapes vs Simulations

“As the modellers wrapped the main water tower asset comprising the tank and legs, we began adding rips, tears, and creases into the tank, being careful not to flatten it. Once the primary level of destruction had been approved, secondary levels were added in the form of panel separation, bent edges and smaller creases, and tertiary animation with snapping tension cables and falling debris. Another modeller was responsible for the girders, twisting the legs based on reference imagery. As the animators picked up the models and the shots started to take shape, we would request further articulation of specific pieces improving the shots every day.”


Simulations were used for the girders falling and bouncing off one another. At the end of the day a decision to take a more traditional blendshape approach was made out of necessity, mainly to maximize the turnaround time required to see iterations. With the blend shape approach they were able to produce a new version every day once the basic pieces were set in place.

Finally arriving at their full CG hero shot brought the challenge of creating water flowing from the cracked tank and under the camera. Detail was added to the water in the form of whitewash foam and aeration, particulate mist, fine volume mist, spurts and splashes. The water interacts with all objects in the environment and causes pallets to shift. Fine ice crystals blowing in the air were added, with snow clumps falling from the collapsed water tower as the tank comes to rest. Hot glowing embers rise from the heated girder in the distance. Nuke was used for compositing.


Tracking Extremis

Method’s team also worked on shots that needed the Extremis effect applied to the story’s primary villain, Aldrich Killian, which demanded very accurate camera tracking and match moving, achieved with Syntheyes and Boujou. They received a cyberscan of the actor from production from which they built and rigged the character. Each shot started with a tight camera track. Once the camera tracks, model and rig were approved they began match moving the actor’s in-camera performance.

“The majority of the shots requiring Extremis in this sequence were medium shots and required full facial tracking,” said Matt. “To further lock the CG model to the plate, our modellers built custom blend shapes, an average of three per shot, which were modelled to match the key points and lock to the facial performance.”


These match moves were important because the Extremis effect does not rely only on a surface component, but also the richness Chris Townsend was looking for comes from the deeper, internal component of the effect that included an internal glow occluded by the bones and vascular system. Therefore, rigging the character for Extremis needed to not only match move the exterior surface, but all the interior vascular systems needed to stay inside the exterior surface.

Plant Regeneration

An interesting animated look that Method was assigned to develop was the rejuvenating effect that the drug Extremis has on a damaged plant. They started with concept paintings made by Method’s concept artist Olivier Pron to show how hot, travelling streams of extremis, referenced from production, rapidly regenerate the plant. The drug starts at the core of the plant and flows up the stem via its internal fibres and into the wounded nub. The plant regenerates when the fibres begin to melt and form a lava-like goo. From under this goo, new growth is formed as the plant cools and ash is brushed away as the new growth emerges.


As the concepts were in the works, effects artist Tomas Zaveckas began building the plant effect using transformation matrices to animate the growth. Once he was able to visualize what the effect should look like he could construct a matrix that performed the desired transformation. He described it by saying, "The plant effect was a transformation matrices heaven!"

Cinesite Powers Up

The final battle in ‘Iron Man 3’ brings together the heroes and villains at night at a floodlit, abandoned marina. This was the primary sequence Marvel approached Cinesite to work on. “We came onto the production just nine weeks before delivery, which is an unusual situation for our team,” said VFX supervisor Simon Stanley-Clamp.

“We had no pre-production background or on-set planning to guide us. The sequence had grown to the extent that the original vendor assigned to it, Weta, still had a number of shots and tasks to complete at that stage. We mainly took care of the sequence opener, as the characters arrive, up to when the shooting starts. All up, between this sequence and some other work, we delivered about 100 shots, nearly all of which remain in the final cut.”

To get quickly up to speed, Simon chose a team he knew well from other projects and kept their approach straightforward. Avoiding any undue technical experimentation, they stuck with what they knew they were good at, using RenderMan for all renders. If their effects library – which is extensive in terms of explosions, smoke and spark elements - already held a suitable FX pass they would use it instead of starting from scratch. They lacked even the luxury of two or three days in R&D.

Marina Environment

“The marina environment came to us direct from Weta, who by then had spent several months building up the assets. They handed over the highly detailed models along with their textures, which was a bonus because we could be assured of maintaining scale and looks. We had a LIDAR scan of the marina set to help during layout, plus a couple of live action plates to match to for the topography. Then, about three to four weeks into our work we received some finals from Weta as well, for a clear idea of the overall look and grade. Fortunately, the Iron Man films have a very identifiable look latch into, such as colour scheme.”


With these raw materials, Cinesite’s team went straight into laying out the marina, getting sign-off on the placement of assets as they went. They had no concepts to work from but used some pre-vis, mainly in the form of early cuts and some bash-composites from editorial. Simon established four views, and they placed their cameras accurately in the scene based on a schematic of the marina. This meant the first pass layouts were based on a true layout. Temporary blocking layouts were produced for each shot and presented to production in the cut and from there, shots could be adjusted to better suit framing for the scene or make a better composition.

Re-purposing and Asset Sharing

For a handful of shots as the gun battle gained momentum, they could re-purpose the live action by nesting it back down into the CG environment, and then on others they did the reverse of that, placing elements back behind the shot plates. In short, the plates were not always intended to exist in the edit as they were captured. They might be reduced, flipped or pulled apart. “These are all standard enough compositing tasks if the team has been on the project from the start. You develop a feeling for composition and how elements should be used. But coming in later on meant some extra layout experimentation,” Simon said.

All communications from production VFX supervisor Chris Townsend came to them via Cinesync conference calls, even the initial brief. There was simply no chance for a face-to-face-meeting. As they got going they were participating in a conference call every day. Simon, amazed at how efficient the production was at handling the work remotely, said, “In total I believe they were working with about 17 or 18 vendors, more than I had artists on my team. Yet the film holds together and never falls into a mosaic of the different teams’ styles.


“Weta’s assets fitted very neatly into our Maya-RenderMan pipeline. Often, asset-sharing needs a lot of back-and-forth communication before the models and textures work properly but Weta’s build and method of hand-over was very efficient,” said Simon.

“At the beginning we had to write a few scripts to convert the file structures and naming, and handle renumbering of textures and assets to something that we could use in our pipeline. However because everything was well packaged up from Weta it became an automatic process by the end, with few manual tweaks. We were also given approved look development turntables, which helped to match our shading and lighting to. Further to that, the shots fortunately didn’t involve rigs or characters to work with, only geometry and architecture.”

Dramatic Lights

The scene is dramatically lit with floodlights that often shine through clouds of smoke and a moist atmosphere. Cinesite put this to their advantage when lighting their work to match this, placing their own smoke elements and atmospheric elements on cards in Nuke and shining digital light through them to give an impression of the path of the follow spots.


Within the marina model came various vehicles such as forklifts, cranes and smaller trucks which the team re-purposed to fill out their own composites, adding and animating headlights and driving them about the scenes. The practical plates were fairly busy with plenty of people and vehicles milling around so they matched their shots to that.

The distant environment and lights came from their matte painting department who produced a massive, oversized digital matte painting based on a practical marina. It was used very subtly, placed into the far distance and shown well out of focus.

Stark Industries

Apart from this big marina sequence the team also completed a wide overhead establishing shot of the Stark Industries complex. A helicopter plate had been captured of a university campus, which Chris Townsend wanted to look more like an industrial science park. They vegetated the surrounding area, removed some of the roads and then built up the park with foliage and trees.

The environment department re-purposed some buildings in the plate by either re-positioning them or placing them on cards to accommodate a camera move. The helipad was built on top of the existing architecture. The grass was matte painted down and the trees were created with Speed Tree software to match the existing foliage, ‘growing’ additional trees to drop in.


They have started using Speed Tree quite regularly now, including on their recent film project ‘WWZ’. It is an application used to create 3D animated plants and trees for animation and VFX shots and integrates with other digital content packages like Maya.

Further shots the team completed included several green screen composites for the in-car shots, all at night using extra reflections on the windows and grading with interactive lights flashing across the faces. Also, the villain Savin’s head, seen closely shaved in the movie, had to be reconstructed following a re-shoot that was completed after the actor had grown his hair back. They had a CyberScan and texture passes of the character, but did not need to resort to 3D methodology for the clean up, achieving the rebuild completely in 2D.

3D Holographic HUDs

Cantina Creative also completed over one hundred 2D and 3D visual effects shots that appear throughout the film. Cantina Creative designed the 3D heads up displays as a virtual graphical interface that Iron Man sees from inside the helmet environment of his armoured suits to receive updated data on his physical condition, weapon and navigational diagnostics and other vital statistics. Their emphasis was on Stark Industries’ most advanced Mark 42 suit, new for ‘Iron Man 3’.


Cantina Creative has previously designed HUD and interface graphics for other Marvel superhero films including ‘Iron Man 2’ and ‘The Avengers’. Due to the engineering advances in these new suits, including Tony’s ability to access them via remote control, the VFX team had to further evolve their designs and graphics to augment Stark’s identity and performance accordingly. The need to deliver shots in 3D stereo added other demands.

“Chris Townsend’s creative input included using a more 3D photo-real, holographic approach as opposed to the way we had worked on the previous films, relying on 2D graphic elements in 3D space. The new designs took advantage of the beauty of the volumetric lights projected into space using optical flares, a design motif drawn from the ‘External HUD’ hologram in the movie, and the textured reflections generated by the light interaction of the HUD widgets from the ambient environment. This meant we weren’t limited to creating graphics that only worked in a text-functional manner but could give them an organic, holographic look,” said Cantina’s VFX Supervisor Venti Hristova.


To handle the stereo workflow requirements in the film, Cantina Creative developed a stereo rig, similar to the one used in ‘The Avengers’, built in Adobe After Effects and then exported into CINEMA 4D for animating 3D objects such as the suit, helicopters, and dimensional navigation graphics seen within the HUD.

Manipulating POV

“Marvel provided us with VFX plates of Robert Downey Jr’s performance as Iron Man in the suit,” Creative Director Stephen Lawes explained.  “His head movements were motion tracked to help define the motion and action beat needed in each sequence and then composited into all our graphics sequences so that we could mirror these moves as accurately as possible.


“CINEMA 4D has stereo tools that let us work in 3D space so that we always knew how far the HUD graphics needed to be from Stark’s head to effectively push his point-of-view beyond the helmet. We imported left and right eye cameras from After Effects into CINEMA 4D and created multilayered lighting and textural HUD effects that could be quickly rendered out into various passes. The software also has some Sketch and Toon non-photorealistic rendering tools that gave us a chance to make artistic adjustments, like varying line weights in space, to communicate more depth and dimension than in the previous Iron Man films.”

For Venti Hristova, one of the team’s major challenges was the initial Mark 42 boot up sequence. “Many of the elements, including a miniature version of the suit and the holographic helmet were generated and rendered from CINEMA 4D. These graphics had true 3D depth which heightened the stereo viewing experience as well as the interactive light qualities, making them both photo-real and immersive.”

Words: Adriene Hurst
Images: (c) 2013 Marvel Studios