In the second installment of VFXWorlds exclusive production diaries, Rhythm & Hues Bill Westenhofer delves deeper into The Chronicles of Narnia: The Lion, the Witch and the Wardrobe. Includes QuickTime movie clips!
This is the second of four installments in VFXWorldsThe Chronicles of Narnia: The Lion, The Witch and the Wardrobe production diaries.
I previously explored our pre-production phase, which involved developing and integrating the software we would need, starting character rigging and development and our responsibilities during principle photography. This part will cover everything else that brought us to final delivery at the end of October 2005.
At the point where this installment picks up, our team is all back at our studios in Los Angeles and our primary focus is on finishing up the significant portions of our pre-production still remaining. We have to finish our character rigging and setup, especially the facial rig for Aslan. We are also heavily into the motion capture and motion editing for our Massive agents and are working on the agent brains so we can start placing Massive into shots sometime in May. While this is going on, we do have our first batch of shots in-house, which are all from Aslans sacrifice scene at the stone table.
Jan. 16, 2005
The final touches for Aslans facial rig are finally being put into place. This is the culmination of work that really started last August. The challenge now, as was then, is that we still do not have a voice from production. The voice really determines a lot about a character and will inform some of the subtle specifics of our facial poses. In the meantime, we have to make do with our best assumptions. Our facial rig is a combination of a shape-based blend system with an additional layer of muscle and traditional deformers layered on top. A shape-based rig uses pre-built facial poses that an animator can select, combine and animate between. At this stage in our development, we have incorporated everything we can until we have a voice and can evaluate whether we need additional improvements.
The process of building the face shapes started last summer by looking at the problem from two directions. The first was to consider the range of poses that a lion was anatomically capable of making. We had an entire room filled with snapshots we had collected from our reference and from the day we spent in a lion cage at Gentle Jungle. From them we found numerous examples of captured moments where a real lion happened to look angry, sad, contemplative and so on. These were useful because they were clearly anatomically possible, and thus on-model, and so were a great way of insuring a sense of lioness in the poses we built. The other way we approached the task was to find a human performer who seemed to exude what we thought of Aslans character. Even with a voice, finding such an example would be useful; without one, it was essential.
Our animation director, Richie Baneham, made the inspired call to propose Gregory Pecks portrayal of Atticus Finch in To Kill A Mockingbird. Atticus powerful, authoritative, sublime persona seemed to fit Aslan perfectly. When Andrew agreed with our assessment, we moved forward with the idea. This allowed us to find still frames from the movie that portrayed through Atticus, a typical Aslan angy, Aslan happy, Aslan proud and so on. With a mockup rig that had been formed with the lion shapes built from the previous stage, we created new poses that had the same feel as the stills of Atticus. Where necessary, adjustments to the underlying rig were made, and in the end our set of performance shapes were complete. So long as the final selection for Aslans voice doesnt clash with the Atticus persona, we will be ready for shot production.
From a technical perspective, we also did a lot of things in the building of Aslans facial rig that improved on issues wed had with past rigs. We had used fully muscle driven facial setups before and know that they produce the most anatomically correct setups. They do this, however, at a cost of being awkward and complex to animate with. The alternative we chose in this case was to use a muscle system to create the blend shapes for a shape-based approach. This would make the rig much easier to work with on a shot to shot basis, while preserving all of the subtle details the muscle systems create. To further improve the process, Wil Telford, one of our creature supervisors, along with lead Aslan facial rigger, Brad Hiebert, devised an ingenious scheme of allowing modelers to free-hand sculpt rough poses in our modeling package which can be done very quickly to get the essences of the gestures. The muscle system would then look at key areas in the posses the modelers had built and determine what muscles needed to fire to achieve the shapes. If a mouth corner was moved, for example, the muscle system would propagate the necessary motion of the cheeks and jaw muscles to achieve a new shape that had complete anatomical integrity. This new shape is what would be incorporated into the final rig.
One issue with shape-based systems is that they often dont interpolate between poses as smoothly as we would like. To combat this, a carefully selected subset of the full muscle rig was added back into the production rig to introduce subtle motion into the skin as the face moved from pose to pose. On top of this, we worked in some special shapes to add major wrinkles to cover situations like a snarl or yawn. Finally, skin harmonics, much like those used on the body of Aslan and the larger creatures, were built in to add vibration to the muscles and skin of the face as he walked.
All of this work has finally been put to the test in by using it to animate to a piece of sample dialogue from one of the potential voice talent choices. The test was successful and is now finished, just in time for us to start incorporating facial animation into several of the shots we have underway.
Feb. 10, 2005
Today I saw for the first time a rough cut of the proposed teaser trailer to come out in early May. Its a great teaser and is sure to create a lot of excitement for the film, but it is chock full of a lot of our shots, many of which we hadnt even planned on starting until June. Perhaps the biggest issue is that several include armies to be simulated in Massive, which we hadnt planned on being ready for full blown simulations until sometime in May. With a copy of the proposed trailer to review, our department heads met today to discuss what options are available since we really want to provide as much for the trailer as we can. The good news is that the majority of our shots are very short cuts, some as short as eight frames. This will allow us to take some necessary liberties to get it done on time. For example, we can get away with simpler Massive agents that just run and dont actually start any of their attack animations, and it will be ok if our final texturing and lighting isnt completely finished on background characters.
Scrambles like these are not uncommon in post-production. Marketing departments need to market their films and naturally want the most exciting imagery to do so. Likely trailer shots are often discussed while principle photography is being shot, but generally other factors such as the order cuts are assembled in editorial, or the availability of voice talent dictate what order shots are delivered to the vendors. When events such as this teaser trailer come along, so long as we have some warning and the production gives us some flexibility to cull out things that really cant be done, we are able to find a way to accommodate them.
Mar. 22, 2005
Through a heroic effort and a lot of simplifications we have our first renders of Massive characters in the trailer shots. The final agent brains that we will use for hero shots are still under development by the team under Dan Smiczek and Lisa Wild, our Massive supervisor and lead, respectively. To accommodate the trailer shots, the team created specific simplified agents who choose a single action as opposed to a weighted selection from the entire motion-tree. An equally amazing accomplishment is that we have been able to transfer all of the detailed prelighting of our hero characters onto the much lower resolution geometries that we use for Massive renders. Our armies will often include tens of thousands of creatures on screen at a time, and it will be impossible to render them if we dont have some way of reducing detail as creatures get farther from camera. Every character rigged for Massive has three Levels Of Detail (LOD): Low, XLo and XXLo. The Low res versions have all of the armor and parts of the hero characters with a lower polygon count. We have even worked out a way to use fur at this level a first for Massive. Since the model is still a topological match to the hero characters, we are able to directly map the texures and shaders from the full resolution prelighting onto the low models. The XLos and XXLos, on the other hand, are much simpler polygonally to a degree that makes direct mapping impossible. A solution to the problem was enacted by artist Jubin Dave, who projected a rendered version of the hero character into a single map that can be applied to these low res models. The result carries a lot of the visual complexity with a much simpler render, and because it is procedural, it can be easily recreated as final adjustments to the hero characters prelighting are made.
Apr. 20, 2005
One of the challenges that I have been most concerned about until now, putting motion-captured pieces of a human together with motion-captured pieces of a horse, has finally been completed with great success. The credit belongs to both the Patrick Runyons motion-edit team and Dan Smiczeks Massive team. The real thing can now replace the hand crafted single cycle for the trailer.
The motion-edit departments primary task is to clean up motion-captured clips from Giant to make them ready for import into Massive. In the case of the centaur, they did specific additional work. With the locomotion actions like walking, trotting and galloping, Patricks team found suitable horse captures and matched them with the closest fit for an upper body. To this, they added procedural overlap to the upper torso placing additional overlap on the head, arms and weapons using the motion editing software Alias MoCap (formerly Kaydara).It took a number iterations and careful synchronization of the two halves, but now it really feels like the upper half is leading the horse body and they are moving as a single creature. These motion clips will be imported into Massive as linked pairs so that the synchronization of the upper and lower halves are maintained.
With locomotion actions it was achievable to link a single upper body with specific horse captures. The problem with battle actions is that doing the same would severely limit the number of choices. It is desirable, for example, to select from a set of sword swipes, blocks and thrusts for a given horse side step. For this reason, the procedural overlaps to extremities that worked so well in motion-edit would need to be built into the agent brains within Massive itself for the combat actions. I have to admit, I was skeptical about how well this would work. We were, in fact, prepared to send selected horse and human actions into animation to blend the integration. But the agent brains ended up working quite well and we are fortunate now to avoid that significant additional workload the required supplemental animation would have had on our animation staff.
May 9, 2005
Today Industrial Light & Magic has been added to the list of vendors working on the film. The final cut from editorial is complete and the shot count has come in at more than 1,400, considerably larger than the original plan. This is not at all unexpected on a film of this size and complexity, but schedules as they are, the production has chosen to off load a lot of the overflow to ILM. Theyll be handling a lot of the leg replacement shots where a centaur, faun or minotaur upper half was filmed in the live-action plate and a CGI lower half has to be added. They will also be adding several full CGI creatures to these shots. The catch is that they will be replicating a lot of the same characters that we are putting into our shots, so the process of sharing what we can has begun. To ensure consistency, we are giving them all of the models, texture maps, turn-tables of prelit objects and motion cycles as we can. There is some question as to the extent that we will be able to directly share some of the data, but we are sending them as much as possible to make the match as successful as we can.
May 23, 2005
The Lion, the Witch and the Wardrobe is going to use a Digital Intermediate (DI) process to handle the final color timing of the film. Digital Intermediates offer a host of new tools to film makers to perfect the color of an image and to avoid some of the generation loss that occurs in a photochemical release print process. One thing that the process sorely needs however, are some firm standards with regard to how digital data is mapped to a film output. This is particularly challenging with darker shots such as our recently completed Aslans Death (AD) sequence. Most major effects houses have developed closed loop pipelines from scan to filmout and have worked in their own lookup tables (LUT) to map their digital file formats to film. With these dark scenes even subtle differences in the mapping of color and contrast can have dramatic results. Up until today, our dark shots have appeared very contrasty on the DI facilitys system. Fortunately, a number of tests with sample scenes have found the appropriate remapping to make our shots look as they should. Unfortunately, a filmout on our camera with this remapping looks washed out so the plan is to place the correction onto our files as a post-process after weve finaled a shot with Andrew.
The DI process offers a tremendous amount of capability in the pursuit of bringing amazing imagery to the screen. With more and more feature films choosing this option, the time to adopt some industry-wide standards is here to avoid these unnecessary workarounds.
May 21, 2005
Today we met with Sony Imageworks, ILM and Dean Wright, the production vfx supervisor, to work out the details of how we will handle shots where multiple vendors are contributing to the same image. The final conclusion is that each shot must be handled in a case-by-case image, with the ideal situation being that companies finish their work in order from background to foreground. Whoever deals with the background characters will animate and composite them into the plate and supply the completed element to the vendor that is creating the foregrounds. In cases where another vendors element falls into the mid-ground, the background will be composited first, then the second vendor will comp their mid-ground characters and then the new composite will be returned to the first vendor to finish off the foreground characters. For the most part, we can handle joint shots in this way, but in the case of a scene where Aslan takes down one of Sonys wolves, we will actually need to share camera tracking data and do a back and forth transfer of animated geometry for us to work out the details of the hit and contact. The trickiest part of the whole process is working schedules to make sure short portions are done in time for the other facility to finish the shot.
All of our imagery is being transferred between facilities in Cineon format for consistencys sake. Weve also conducted scanner comparison tests to make sure there wont be any issues in dealing with each others data.
Jun. 15, 2005
At long last, we have a voice for Aslan. Liam Neeson has been chosen and were receiving cut dialogue to begin working into our shots. The best news is that his mannerisms work perfectly with the facial rig we created based on Atticus Finch. First passes at dialogue animation are working very well.
Jun. 24, 2005
With the exception of a few shots of the girls riding on Aslans back, we finally have all of our shots in house to work on. In a perfect world you get shots in just the right order to optimize the way you assign artists to given tasks. It is generally best if you can keep the animators working on the same characters throughout a production. You also try to keep sequences, and specifically shots that are identical in terms of camera angle and lighting with the same set of artists. The needs of the visual effects vendor in this regard are rarely the deciding factor on the order of what gets turned over, however. More often it is the order shots are filmed and thus the order that sequences that mature first in the edit that determines this. As a result, you learn to be flexible with artist assignments and do what you can to keep them all working as effectively as possible. Now that we have received are full complement of shots, we can shuffle current assignments to try our best to get back to this optimal workflow.
Aug. 5, 2005
Even though weve been finalling shots with Aslan in them for about two months, we have still been making minor refinements to various aspects of him, especially tweaks to paws and parts of his mane. When we fix something significant, we will occasionally unfinal a shot and re-render with the newer look. Today, however, the last of these changes have gone through and we are considering his model locked. This is the end of a process that started in March of last year with the arrival of the maquette from Weta Workshop. The work done by the team on this character has exceeded what I thought to be possible with a CGI creation. He is by far the most complex hero character Ive ever been involved with.
Aug. 24, 2005
Today we delivered a whole host of shots to the client for a temp screening that will be coming up in a few weeks. The significant step is that these now have, for the first time, an army of Massive agents incorporated into them. Weve already had great success with the package on individual shots such as the crowd scenes around the Stone Table. But this was the first time that weve been able to turn out a large number of shots at once across an entire sequence. Its funny how you get used to the empty vistas, or more recently, empty vistas with a small number of foreground hero CGI characters. It wasnt until this pass that the battle took on the epic scope that it was intended to be. It was a long road of software development, testing, motion-capturing and brain-building, but the investment is finally paying off, and doing so far more quickly on a shot-by-shot basis than we expected.
Sep. 9, 2005
Today we presented the last of our hero character animation for review to Andrew. There is still a fair bit of technical cleanup to go, but as far as the performance is concerned, Richies team has completed its task. The technical cleanup that remains is primarily focused on issues that affect our technical animators such as bad deforms or feet that arent quite on the ground as they should be. Tech Anim is the department charged with animating fur as it blows in the wind, cloth dynamics, skin harmonics, etc. Dante Quintana has been leading the team, and now that principle animation is finished, their workload has increased dramatically. It utilizes a lot of physical simulations and, unfortunately, imperfections that you might have gotten away with in the character animation stage are often revealed here. A subset of the character animation team will continue to the end of the job to make these corrections.
One of the biggest issues weve had to deal with in the Tech Anim stage is wind. While on location in New Zealand, I would often jokingly call this film, The Lion, the WIND and the Wardrobe, because there was rarely a time when we werent being buffeted by strong gusts all day. Even though on screen, the effect seems far tamer than what we experienced, we still need to make sure our CGI characters are affected to the same degree as the other objects in the frame. Our software team has given us two types of wind control: Dynamic Wind and Pelt Wind. Dynamic wind controls handle the broad scale motion of wind as it bends the large mass of Aslans mane around his form. Here, hair-to-hair collision detection is also determined to help maintain a sense of mass the in hair volume as it blows. Pelt wind allows individual hairs, especially at the tips, to blow independently. This is most evident in a close-up when there is a slight breeze blowing the frizzled ends of the more wispy hairs.
Tech Anim is also responsible for all of the cloth hanging from armor, the flags in the hands of the massing armies and the tent flap that moves to reveal Aslan when he emerges to greet the children. We are using our custom cloth solver for this and have integrated Massive output with it to handle cloth simulations for our Massive agents.
Sep. 11, 2005
Our supervisors and I watched the rough cut of the film tonight, and it was great to see our work and the work of Sony and ILM starting to come together. Its always a boost to the teams spirit, especially now that hours are getting long and the work pace more hectic, to see that you are working on a project that has the potential to be truly spectacular.
Oct. 28, 2005
After nearly two years of work, from the moment we were awarded the job, through shooting, motion capture, hero animation and final lighting and compositing, we delivered our last batch of shots for The Lion, the Witch and The Wardrobe. The final weeks were quite intense. Our sequence supervisors Dave Lauer, Art Jeppe, Mark Rhodal and Raymond Chen and their teams all worked hard to make the final tweaks and adjustments to deliver some truly exceptional work. The last week, in particular, took a lot of coordination and back and forth reviews with ILM and Sony, as we incorporated their work into our shots and vice versa. In total, we delivered 380 finished shots to the production, touched about 460, and if you count Massive agents, animated the motion of more than 460,000 characters. Bob Mercier, our digital supervisor, has just informed me that we also completed more than 200 years of render time with our processor farm.
Being a part of the production of this film was an extremely rewarding experience for me. It was exciting to have the opportunity to work on a story Id loved since childhood; and then to watch this talented group of artists and technicians more than do it justice with exceptional imagery and animation was a wonderful experience. I want to thank Andrew Adamson, Walden Media and Disney for giving Rhythm & Hues the chance to contribute to what is certain to be a cherished classic.
Bill Westenhofer is the visual effects supervisor for the Rhythm & Hues team on The Chronicles of Narnia: The Lion, the Witch and the Wardrobe. His numerous credits as visual effects supervisor include: Elf, The Rundown, Men in Black 2, Cats & Dogs, Stuart Little I and II, Frequency and Babe: Pig in the City, the latter a nominee for Best Visual Effects by the British Academy of Film and Television Arts. Starting as a technical director with Rhythm & Hues in 1994, Westenhofers lighting and effects animation were featured in Batman Forever, when he first worked with Narnia director Andew Adamson, and numerous television commercials. Westenhofer quickly rose to CG supervisor for Speed 2: Cruise Control, and continued in that role for Spawn, Mouse Hunt, Kazaam and Waterworld.