Weta Digital Conference in Milan
Hey on 26.04.2004 Alias Wavefront, presenting its new software Maya 6.0, organized an event in Milan with Shawn Dunn from Weta Digital. He showed the making of The Return of The King, being involved in it for almost a year. I was there, and I was lucky enough to interview him! He is really friendly and forthcoming, even if he couldn?t say anything about ROTK extended or King Kong! However, before passing to the interview and the report, let me thank Oronzo Cilli of the Italian Tolkien Society who indicated to me the event, and to Alias which provided to me the entrance.
A brief interview with Shawn Dunn, Weta Digital Animation Technical Director.
-When did you join Weta digital crew? How was the atmosphere there?
I joined Weta Digital in March of 2003, it was during the postproduction of Return of the King
-Are you involved in the next Weta Digital projects, as KK?
Yes, I?m still working with Weta, and I?ll be there for another couple of years, so I?ll work on King Kong and the other projects.
-When you were working on Return of the King, did you feel as you were working on a eleven-oscars movie? Weta Digital must be proud of its 3 oscars in 3 years!
I think it. Working on RotK was an incredible experience. And I?m really proud to be involved into the project. When I got there and started to see the scenes that people were producing, I began to have the sense that I was doing something really, really amazing and huge. And you can look at it: look some of the shots, later during the conference. I was involved into that, I followed the making of that scenes step by step, but every time I see that movie on the screen, it makes me goose bumps! It?s pretty amazing being part of that team.
-What about your 2 thousand processors? Is it true they are actually onleasing? Who is renting them currently?
I don?t exactly know who is currently renting them, we have this huge rendering-wall, which is divided in two parts, and I know that we are currently leasing one of them. The other part we?ll use for our current projects, like I, Robot or Rotk EE. But be sure that for King Kong that render-wall will be really busy?
-What an average day involved during last year of postproduction?
Well, for ROTK I was the Animation Technical Director. My role was primarily to support the Animation Team, so I was building tools for them, I was helping them to solve the problems, making work flow? We worked a lot, many times we thought it was impossible. Some days we finished a scene and had to begin with two, and my work was to keep them working well.
-a very important question. The leader of the Mumakil that?s killed by Eomer. Was he totally real? Or totally animated?
I guess he was real only in really close ups. Anyway, it?s totally photorealistic.That?s real during the really close-ups. When he plays the horn, he?s real (p.s. in fact, during the conferende, we saw a live action plate with an actor in a bluescreen studio playing the horn). But in wide scenes, and when he?s shot by the spear, that?s completely cg. That scene, when he falls down from the mumakil, and the mumakil itself hurts another olyphaunt, is one of my favorites scenes.
-Did you animate the olyphaunts totally by keyframing?
Yes. It?s not something we can do motion capturing for, it’s hard to get an elephant to stand into a studio and so on?!
-speaking of the Extended Edition of rotk, we know that the new scenes have just been finished. How many new visual effects scenes have been added?
I can’t talk about that, sorry. How many more S.E. scenes? I can only say that they are a lot. When I left them they said to me: you can talk only about some things?these aren?t into that things. Just because isn?t released yet, you know…
-what about king kong and Evangelion?
All I can say is that we are working on King Kong?sorry
-No matter. Thank you very much and good job!
The conference was pretty amazing. He projected many parts of the movie, and explained how they did them. Many scenes were ?dissected?, so we could see how was the clean plate, and then, layer after layer, the finished plate compositing itself. It?s really amazing to realize how much work is behind this trilogy, especially the third movie. I?m serious: It?s really impressive the huge amount of work they had, and the passion and the devotion they invested during this years.
It began with a reel of clips from rotk, before that he spoke. The clips were the most important visual effects scenes of rotk, and obviously involved Minas Morgul, the trolls, shelob, the mumakils, the monsters pulling Grond, the nazguls and the Fell Beasts, Legolas on the Mumakil and? Gollum.
After that he explained the difference between Weta Workshop, which makes physical effects, and Weta Digital, which makes visual effects. He said that they are quite different, and even if now they work separately (W.Workshop?s working on Narnia and other movies, W.Digital made some scenes of Van Helsing and is working on I, Robot and the extended edition of Rotk), they?ll reunite for King Kong, which is currently on pre-production (Weta Digital is currently developing the technology for Kong).
For RotK Weta Digital had a huge amount of work, and it was a run against time.
They finished the last scene only one week before their last deadline. It had more than 1,527 visual effects shots. Whereas FOTRhad accumulated 7.5 terabytes of online data, and TTT 30 terabytes, ROTK amassed 72. They had 600 workstations and increased the amount of processors of 1200. He said that if they would processed the movie with only one of their 2,100 processors, it would take 11 millions years to render all the three movies. He showed, projecting the display of his personal computer, Weta Digital?s pipeline: a diagram that organizes all the steps of the postproduction of a movie. This is linked to a sharing of data that has to be really functional and quick, and to an organization of the data really tidy. Most of the creatures were modeled and animated using Alias? Maya, that software
was involved almost in every step of the pipeline (modeling, creatures, cameras, animation, environment?).
To make a creature they would do many sketches as first step, then sculpted many models in clay. Than Peter approved one, than they scanned it with ad high-resolution scanner by hand (unto 60 millions of polygons). After that they generated a low-resolution model and compared it with the hi-rez model. Between them they did many displacement map. The ending was a low resolution model easy to animate, but working in high-resolution (with a lot of details) when rendered. Gollum was animated in many different ways. They usually motion captured Andy Serkis: as we know, he acted on location with the other actors, and then re-acted all in a motion capture studio. Other times it was ?rotoscoped? on Andy Serkis, and other times was totally animated by key-frames. The face, for instance, was inspired by Serkis? acting, but was totally key-framed, using a library of facial expressions too. For RotK Gollum was really improved. He showed us that amazing scene that?s when he speaks with himself sleeping. He said that they improved the lips, as you can see in that scene they are stickier on themselves, wet, and more realistic in the dynamics. They improved the eyelids, the way they move and flow closing themselves; and the way the muscles swell: he said that they use a plug-in they wrote for Maya to make the muscle system of LotR models. For RotK they also worked on Gollum Model making that thinner and haggard, as the character evolves in the book. So it began even more difficult to transfer Serkis? movements on Gollum Model, because of the difference between the proportions.
Gollum was also the only creature that had close-ups needing an interpretation and some kind of ?presence?. For RotK, they improved their library to more than 600 different facial expressions.
He opened the Facial Animation System, and showed us some of the facial shapes. They used to combine and melt different expressions to make a facial animation, using keyframe.
But also every part of the face was animable separately: the eyes, the tongue, the lips, the eyebrows? The wireframe model of the head wasn?t a low-rez model, but a medium-rez, because of the complexity of it. Obviously the hi-rez details came from the displacement maps during rendering. He showed us many interfaces Weta Digital uses to organize the pipeline and the data. First of all there was the Weta Digital Manager, a database of models or scenes divided by movie. On the left there was the LotR section, the Van Helsing section, the I, Robot one and?the King Kong one. Obviously he opened only the LotR one, and I began to drool thinking that there was kept some of the early Kong CG models!!
Then into the Setup he showed us many models they made for LotR: Ents, gondorian soldiers or riders, digital doubles of the actors, the King of the Dead?s body?and the King of the Dead?s head! Thanks to that ?organizer? and that interfaces, every animator was able to search and obtain every kind of data or model he looked for. Every time someone finished to work on a shot, ?published? it on the Weta Digital Animation Publisher, so the shot could go to an higher step of processing.
The Fell Beast Model was really complicated to animate, because of the neck and the tail which were long and supple. Every time they moved them, they had to pay attention the dynamics wouldn?t make the model hurt something in the scene: a soldier, a building?
Shelob was really complex. They decided to change the position of the eyes (that in a normal spider are located in another part of the ?head?) because they wanted her to interact with Sam during the fighting. In that way she could see him during the fight. He said that when she rolls up Frodo in her web, that is totally CG.He showed the clean plate, which was shooted in studio with only stones and a painted background, and then they added the CG Shelob and Frodo. He said the web was really complex to animate. And the fight scene is almost totally CG, unless in close-ups. When Sam hit Shelob with his feet, he is real, Shelob is digital, and Sam?s feet are digital too!
The strange triceratops-beasts that draw Grond were something they go really proud. They appear only for few shots, but for that they produced sketches and models and worked a lot! All is digital there, Grond is a real miniature only in some close-ups. The Trolls were 6 and were different each other. They were really high-detailed, thanks to the displacement maps. The problem with them was that because of their size, they had to control everything worked ok, every movement they did. He showed us many scenes in which the Trolls destroyed everything, squashing people and launching soldiers, and said that some animators liked that ? really sadic!
Then he showed the destruction of Barad-Dur: that was a model, too, and they literally destroyed it. The pieces of the tower were animated using Particles Dinamics in Maya. He showed us their Eye of Sauron Model: it was made of 60 melted layers that formed a 3D texture, so that it was transparent and three-dimensional. We saw how they made the enormous ?hole? in the ground, as the tower explodes, and he said that it was all animation, not a miniature. He added that on the Black Gate we can see the Trolls of The Two Towers fall and die: they animated them, but almost nobody see them!
Showing us the enormous, magnificent miniature of Minas Tirith, he explained how they realized a digital model of it, but used only for pre-visualization and 3D environment to track Gandalf digital double during the raising to the citadel. He said that almost all the scenes with Minas Tirith were made with the miniature, because is more realistic: their motto was to do the most possible with real effects, because are cheaper and more real than digital one!
He showed how they carried out that amazing scene that?s the descent of the Nazguls on Minas Tirith: we saw the clean plate, that was made with motion control (that?s a technique indispensable for this visual effects shots) with the miniature and nothing else, then came the 3D environment, so they could work on the shot adding the ground and the mountains and other elements, then came the backgrounds (amazing 3D matte paintings and mountains totally generated, not with photos), then the animation of the Fell Beasts and the Massive Agents on the citadel terrace, then the clouds? Really a fascinating process.
He said that the catapults are all completely CG, unless in some close-ups where they used miniatures. The Siege Towers were digital, too: full of Massive Agents. He showed us some of the common problems they had with the animation: in a wide shot of the Siege Towers, after rendering it all the night, they came in the morning and saw the Siege Towers literally spinning around, with the Agents flying miles away! He also showed a joke the animators did: they animated a scene with some orcs shouting and bringing on a catapult one drunk orc (he took an X bottle in a hand!!), throwing him and exulting! He-hee, quite mad..
For the Horse Massive Agents they had an enormous database with data of the motion captured horses (we saw one of them: it was quite strange to see it full of sensors, but it wasn?t sad, I guess!). They did some Models and then Massive randomized them and produced thousand of horses.
The sequence with the eagles fighting the nazguls was made by two really skilled animators: they made it in only two weeks.
Thanks to the pre-visualization, Peter Jackson could strictly control all the postproduction process. He knew very well what he wanted, and using the 3D pre-viz (which were largely used in RotK) he could visualize perfectly all the complex scenes. Saying ?I want the final shot completely identical to the pre-viz? he could obtain exactly what he wanted, because after all he compared them?
In the end, he showed us the Mumakil Models and said that they were different each others, with different sizes, tusks, colors and textures. The towers they brought were 3D Models, but they didn?t animate them: they followed the Mumakil dynamics. They filled them of Massive Agents, but also of Live Action actors, as he said to me during the interview. You can see the photos of the event here: let put them on your scrapbook, if you want! ^__^
Andrea F.Posted in Old Special Reports on May 3, 2004 by xoanon