In November 2016, Alexis Wajsbrot and Jonathan Fawkner talked about the work of Framestore on DOCTOR STRANGE. And in April of this year, Kyle McCulloch explained to us about his work on BEAUTY AND THE BEAST. They are now talking about the last act of THOR: RAGNAROK.
How did you and Framestore get involved on this show?
Alexis: Framestore now has a great relationship with Marvel, having worked on several of the studio’s movies such as IRON MAN 3 / CAPTAIN AMERICA / GUARDIANS OF THE GALAXY (VOL 1 & 2) and DOCTOR STRANGE. Marvel especially loves the work Framestore was able to achieve on the characters such as Rocket, Baby Groot and even Doctor Strange’s Cloak of Levitation, so it was no surprise that they ask us to collaborate one more time on the third act of THOR: RAGNAROK as there was a lot of character work involved.
Personally, I joined the show at the very beginning of the post production as VFX Supervisor to team up with Kyle McCulloch to deliver the massive amount of work, as we initially were suppose to work on both act 2 and 3. At the beginning I was mainly focused on Act 2 and the Sakaar work, until act 3 got even bigger and we put all our crew on it.
How was the collaboration with director Taika Waititi and VFX Supervisor Jake Morrison?
Alexis: Taika Waititi is really relaxed and makes the reviews a very fun and collaborative process. He does not hesitate to mimic the actions, make jokes, and bring a lot of ideas to improve the images. His main concern is always about rhythm of the animation, to make sure all the comedy aspect is as strong as possible. I remember a shot where Banner falls onto the bridge. Taika made us remove some of the cloth work to make Banner fall even stronger and funnier. It does work, everyone was laughing hard at that moment in the screening. Taika is also very focused on the symmetry of the shots, so we made sure all our assets were symmetrical and made the action as centered as possible, it definitely gives a very graphical and stylised comic book style to the shots which look great.
Jake is a very experienced Marvel supervisor, and he knows the THOR universe especially well as he supervised the first two movies. His input was incredibly valuable and we managed to make Asgard even better than in the first movies. We had 2 cineSync reviews a week with Jake and it was a great place to collaborate and emulate ideas to make the shots better.
What are the sequences made at Framestore?
Alexis: It was mainly act 3; the fight on the rainbow bridge with all our super heroes teaming up to fight Fenris, an army of D-guards, Hela and Surtur. We also worked on the last sequence of the movie within the spaceship.
How did you organize the work at Framestore and between you two?
Alexis: Very good question, as it was mainly one sequence with the same assets. It definitely was a challenge to split the work between Kyle and I, but also was a challenge within the team. The way we solved it was to split it chronologically; I supervised the first part of the sequence including the Asgardian crowd, the commodore crash and fireworks, Loki’s arrival with the build of the Statesman (a giant ship), the D-guards fight, Uber-Thor (including the lookdev of the lightning bolts) and the Fenris vs Hulk fighting on the bridge and into the water including the asset dev.
Kyle supervised the second part which included the lookdev of Hela, Asgard, the Rainbow bridge, Korg, Miek and especially the Mega Surtur beat including the asset lookdev.
Team- wise the team was split between us 2, and we tried to limit the crossover so artists were not pushed with 2 very tight demanding requests. The team was composed of 3 very strong CG supervisors as well as some of the best Lead TDs at Framestore.
How did you approach this massive final act?
Alexis: I was scared and excited by the challenge! Working on a third act of a movie is always tricky as it’s the pay off and working on a Marvel third act is even more of a challenge! It’s the final big battle between all the heros and the baddies and it’s the latest part to be locked in the edit so we need to be able to respond very quickly to changes up until the last minute.
The first part was to build and nail the look of the assets. So we did a full round of 2D concepts to start the discussion with the studio and get as many answers as possible. With the answers in hand, we tried to get all of our assets in shots as soon as possible to get everyone at Marvel involved for the best possible feedback.
The sequence have a lot of full CG characters. How did you design and create these characters and especially Hulk?
Alexis: We start with 2D concepts we receive from the studio, and then we do a batch of concepts using Framestore’s Art Dept. As soon as we have some traction, we move into 3D, and we try a rough model / texture / lookdev to show the character in shots or place them into a vignette (small action sequence, where we see the character in motion), and depending of the reaction we either go back to the 2D concept or just update / improve the model. All the new characters were very challenging; Korg was made of rocks and still needed to speak and express emotion, so he was both an artistic challenge as well as a technical challenge to get the rocks to transform and not deform. Hela’s costume and headdress was really challenging to nail. Miek has been more straightforward and our first vignette was really liked by the studio, so he is very similar in the final movie as our first animation test. The look of the D-guards changed drastically during the movie, from being more animalistic to looking more like a human zombie.
It was great for Framestore to have a stab at creating THE HULK! We were all incredibly motivated by the challenge as it’s historically an ILM asset. We received a model from ILM, and we had to ingest it into our pipe. We had to start the rigging / muscles and facial rig from scratch as we use proprietary tools. We really pushed our muscles and skin simulation ‘Flesh and Flex’ tool to get the best possible results. We also redesigned our shot sculpt pipeline to make sure we had the finest detailed wrinkles and veins and could be very reactive to the studio feedback. We had to groom him dry and wet depending on the shot. We also pushed our shaders using a new ulti-subsurface system that scattered light through skin, muscle and bones to render the best Hulk possible.
Fenris and mega Surtur of course were some of the most challenging characters, which leads me onto the next question.
Which CG characters was the most complicate to create and why?
Alexis: For my part it was Fenris. As with any new character, we need to agree on her look (it’s a « she »)! is she big ? how big is she ? in compare to Hulk? to the bridge? Is she rabid? is she half dead and more like a zombie dog? what colour are her eyes? We went through a round of 2D concept first where we tried to nail her look. In the first version Fenris was suppose to transform from a normal dog to a giant wolf, so Fenris looked a lot more like a husky as we shot with a Husky named Dickens on set. After a few iterations, we all settled on a giant black wolf, and his level of threat would be driven by his posing and facial animation.
The other main challenge was the muscles rig and flesh deformation, and finding the right balance between not having him completely rigid but still selling his scale by not having Fenris’s skin flapping too much.
Her groom was definitely challenging to achieve, and Fenris has a different groom variant depending how wet she is. First we aimed at making Fenris look photo real and we didn’t focus on her scale too heavily. We used a lot of reference of real wolves; their groom, their facial expression, how they move, we also had a real scan of Dickens (the dog used on set) as well as a wolf, which was great information for skeleton proportion.
Regarding the scale challenge, we just made sure they were a lot of variation in her groom, in term of clumping, greasiness, bold patches with scars, some dirt/ small debris attached to the groom etc. We also made sure with Jake Morison and Taika Waititi that we were always using low angled camera either at human height level or lower. We always tried to have an element of the bridge or a D-guard in frame for scale reference.
Can you tell us more about their rigging?
Alexis: All our rigs now have very complex skin and muscle simulation rigs within them. Thor / Hela / Fenris all used our latest solver Flesh and Flex technology. Korg had one of the most challenging rigs, as he is made of tiny rocks. We spent a long time developing his rig to get the rocks transforming and not deforming / to get them to collide between them. It was especially tricky for the facial rig as we had to model and layout the rocks a certain way to get the best results.
Fenris and Mega Surtur are massive characters. How did you manage their animations challenge?
Alexis: In a very different way for both characters. Fenris is big but not planet-size like Surtur. We started with the previz animation, and then studied wolves and dog motion. We did do some run cycles and made them slower to give more scale and weight. We also played a lot with the camera to make sure they were low angled, Fenris could get out of the frame, and we added some camera shake when we are close to her. Regarding Surtur, it was a very different challenge as Taika wanted Surtur to move fast even if he was huge.
Kyle McCulloch: From the start, Taika felt quite strongly that even though Mega Surtur was planet-sized, he shouldn’t appear to move slowly. This was a big challenge for us, as one of the key ways to convey scale in a huge character is how they move relative to the world around them. From the beginning, the filmmakers were quite attached to their previz, and our animation team dove in on building from those performances, and bringing new ideas to the table on how best to sell the massiveness and gravity of the Titan of Fire.
The three elements that proved key to Surtur were animation, shading, and simulation, and we started parallel work on all three aspects. The animation team iterated repeatedly on different variations of performances, working with Jake and Taika to find exactly the right balance of threat, speed, and scale. We looked at sword fighters, wrestlers, and bodybuilders for inspiration on the shapes and poses we wanted to use on Surtur, always keeping in mind his new-found strength and power.
Can you tell us more about the fire and lava of Surtur?
Kyle McCulloch: The brief for Surtur included references to the surface of the sun, plasma, lava, and fire. The lookdev team spent months partnering with our shader writers, experimenting with a myriad of ways to have something feel huge, somewhat transparent and refractive, while still seeing a visual complexity and surface. In the end, much of the look of Surtur’s body was a comp-led combination of a set of passes that isolated fresnel elements, refraction, muscle volumes and skeletons, as well as a cracked, lava-flow surface.
Lastly, the simulation team developed a variety of volume and simulated elements to round out the look. The FX team provided various volumes and point clouds for the shading team to drive more complex lighting models.
Thor can manipulate the lightning. How did you created this FX element?
Alexis: I would say the main thing that sold Uber Thor was to go full CG for Thor in order to get the correct light interaction between the lighting bolts and his body. We did model / groom / shade a very high res version of Chris Hemsworth that could work in both closeup and far away. We also created a muscle rig especially for the arms and a cfx rig for his cape and costume.
For most of the shots, we received plates of Thor fighting stunts where he is dressed in a mocap suit on blue screen. We body tracked Chris even it was going to be mainly CG, as we wanted to keep his face and facial expression. Once body tracked, animators would slightly change his body performance to better interact with the D-guards, and sometimes replace a leg to remove the feel of the wire work. We simulated his clothes and cape, using the plates as reference.
Regarding the lighting bolts themselves, we used Tesla coil references for the motion and a Houdini setup where we were able to art-direct the density and the speed of the bolts on him. Thor was also able to emit a lighting bolt to kill D-guards and even call lightning from the sky. The lighting bolt was generated using point clouds, we also generated a volume around to emit light which was less noisy and faster to render for the light interaction on both the environment and Thor. We generated a plasma (smoky) pass that was lit by the bolt to make it more detail. When we see Uber Thor’s foot contact the bridge we also added some extra 2d bolts inside the rainbow bridge emitted from his foot.
How did you design and create the huge environment of Asgard?
Alexis: We started using the assets of THOR: THE DARK WORLD, and did have to upgrade them; texture and shade them. We built around 9,000 buildings, mostly laid out by hand using our instancing system of assets called fShamble. The team sculpted mountains based on references of the Norwegian Lofoten Islands, and refined the most distant panoramas as digital matte paintings. Building Asgard ended up being the easy part, the most challenging part was to nail the time of the day and the weather the sequence was going to take place, taking into consideration that we needed to integrate plates of Asgardians shot with a pretty flat lighting. We pretty much tried every configuration before settling for a stormy day light sky with pockets of light the first trailer and then for the movie we ended up with a very low sun, almost golden hour, to justify the Uber Thor and Mega Surtur beats.
Can you tell us more about the crowd creation and animation?
Alexis: We modeled 10 heros Asgardians; 5 male and 5 females and different variants of D-guards. The Asgardian crowd was created in Massive using a mocap clip. We did a lot of extension work for the crowd, and on some shots it’s Full CG; we actually went quite close up into a top down shot.
Regarding the D-guards, we had some sessions of motion capture, the very first mocap performance that we captured was a lot more animalistic in character. We then created a first pass blocking animation on every shot matching the intention of the previz. From there we did new mocap sessions in the Framestore mocap studio, where we directed specifically for the shots and made the animation a lot more human. We are using a combination of crowd D-guards for background using mocap clips, and hero hand-animated D-guards in the foreground.
The final act ends with the complete destruction of Asgard. Can you explain in detail about it?
Jonathan Fawkner: We wanted to destroy Asgard in a series of layers to make it a more visually arresting detonation, but also in a sudden and surprising way, given that it is done as comedic counterpoint, which means we had to do a lot in a short amount of time.
Firstly as Mega Surtur has already plunged his sword deep into the crystal heart of the world, we started with a semi primordial landscape with flame and cracked mountains spewing lava, tidal waves and waves of flame, all at macroscopic scale, as seen from space.
We decided it would be nice to have all the water and clouds on Asgard dispersed with a shockwave, so that all went first. We simulated particles to blow the surface water and waterfalls of the disk of Asgard and we used motion vectors to dissipate the matte painted clouds in Nuke.
Various shell like energy fields were added and then the master simulation consisted of crystalline and rock geometry, with separately advected fluid and dust passes, all emanating and silhouetted by what is left of the once bright heart of an ancient realm.
How did you manage the rendering of so many elements?
Alexis: A lot of optimisation and A LOT OF CPUS! An average of 20 000 cpus. If we were to render THOR on a single cpu it would have take 3921 years!
What is your favorite shot or sequence?
Alexis: I love the sequence of Fenris fighting Hulk into the water, but if i had to pick one shot, it would be Thor with his lightning power jumping into a pile of D-guards; a very iconic shot.
What was the main challenge on this show and how did you achieve it?
Alexis: There were lots of challenges! Integrating the Asgard crowd shot on a stone arch bridge onto the rainbow bridge and making Uber Thor full CG for the lightning interaction are up there with the most difficult. But if i have to choose one, it would definitely be the Hulk and Fenris fighting in rapid water.
We received previz of the whole action by The Third Floor, that of course was being updated through the editing process. First we started with Fenris and Hulk both roaring and fighting on the bridge as some of these shots were present in the first trailer. The biggest challenge was definitely the water simulation lead by Julien Legay, one of our FX TDs on the show. The camera from Taika was incredibly close to the rapid water and we have 2 giant creatures creating extra splashes. It was really hard to clear the camera but still make it feel like natural water simulation. It was the same for Hulk and Fenris; the first simulations were very quickly obscuring our heros so it was all about finding the right balance between making the water sim big but still showing the performances. We did push the limit of both our internal water solver flush and our rendering engine arnold to get the best water sim. We did re-think completely how we were mixing spray and foam with the meshed water using some aeration passes to create white water for better integration.
On the last month of production we had to deliver 2 very challenging close up waterfall shots where Hulk is falling before grabbing a rock of the cliff. They went surprisingly smoothly, and I am very happy to say that Framestore is ready to handle bigger water shows!
What is your best memory on this show?
Alexis: The last week of delivery! We delivered 195 shots that week out of the 459 we delivered in total. It’s usually the trickiest and the most stressful, but everything clicked, we even manage to deliver a very complex shot of Hulk into the waterfall in just 2 weeks, all the asset work really paid off. The atmosphere and motivation from everyone involved was great. It really felt good to be working at Framestore, like Heimdall say in the movie Framestore ‘is not a place, it’s its people!’
How long have you worked on this show?
Alexis: A year
What’s the VFX shots count?
Alexis: 459 shots delivered
What was the size of your team?
Alexis: I looked at the stats and we peaked at 309 people, and the project took 38 052 man days.
What is your next project?
Alexis: I can not say it yet, but I can say that as I got married during the production of THOR, I am currently enjoying a well deserved honeymoon with my wife!
A big thanks for your time.
// WANT TO KNOW MORE?
Framestore: Dedicated page about THOR: RAGNAROK on Framestore website.
© Vincent Frei – The Art of VFX – 2017