| |
Making of Stay Alive
Interview with members of the visual effects crew at Pixel Liberation Front about creating 140 shots for the horror movie Stay Alive.
March, 30th, 2006, by Raffael Dickreuter
 Kent Seki, Visual Effects Supervisor
Tell us about your role as VFX Supervisor of the film and what kind of work this involved from start to finish in the project
The Visual Effects Supervisor acts as a conduit for the director to help achieve his/her vision for the show. In the early stages, this involves listening and acting as a sounding board for the director. Then you work closely with the Visual Effects Producer to plan the Visual Effects. When photography begins, you represent the VFX department on-set to ensure the VFX for the shots are achievable. After principle photography ends, you oversee the final effects for the show.
In the case of Stay Alive, the VFX Producer was the very capable Gayle Busby. We worked together to plan the shooting of the VFX to fit within the shoot schedule and the budget. I was on set throughout the five-week shoot. Then, I went back to PLF and worked with our internal team of over 30 artists to complete 140 shots. In order to accomplish this task, each artist had a significant impact on the production. The project required everyone involved to wear several hats. Technical Directors Andy Jones and Brad Friedman refined the proprietary pass rendering pipeline that we began developing while working on “Sky Captain and the World of Tomorrow.” We created over eight fully digital environments, fifteen characters, and over 100 props and effects for “Stay Alive.” Back at PLF, I was lucky enough to work with internal producer Tanissa Potrovitza. Meanwhile, Busby stayed with production and oversaw an additional 40 shots from Mark Dornfeld at Custom Film Effects (CFE). I can’t stress enough the value of working with good producers. They keep you on schedule, on budget, and organized. In the case of “Stay Alive,” both Busby and Potrovitza did a fantastic job.
Talk a bit about your on set supervision and what work and planning it did involve for the post production process
On-set I was responsible for making sure that we had everything we needed to successfully complete each VFX shot. This meant working with everyone from the Assistant Director (AD), the Director of Photography (DP), Props, Special Effects, Stunt, Hair and Makeup, and Wardrobe. It was a great experience. The majority of shots were screen comps which involved making sure we had a green screen up with tracking markers. In addition, you make sure and record as much data as possible about the shoot. This ranges from camera measurements to digital reference photographs. All this data can be extremely valuable when trying to recreate set conditions in CG well after a shoot has been completed.You always have to balance your anticipated needs in post-production with the immediate needs of the set and the Production’s need to stay on budget and on schedule. This involves the ability to quickly think on your feet and creatively solve problems efficiently. You also need the ability to negotiate with the various departments to try and make the VFX shots as painless as possible while taking the least amount of their time as well.
What creative decisions were made about the look of the game and characters to give a next-generation feeling?
One aspect of the visual effects that proved to be a formidable challenge became the look of the game itself. We wanted to achieve something that could be qualified as “next generation” but yet still believable. To achieve this directive, we incorporated well-known and used CG film techniques into game style scenarios. Since film benefits from rendering while most games rely on real-time graphics engines, we wanted to integrate the more time-expensive rendering assets into our faux gaming world. For example, we used 3D motion blur, ray-traced reflections and shadows, area lights, and volume lighting effects. All of these techniques are making their way into games now as the graphics cards and systems become more robust and real-time shaders gain sophistication. In addition, we wanted to go for a stylized look relying on a heavier application of ambient occlusion. We added cloth simulation and hair rendering. All of these techniques have been creeping into gaming gradually as gaming and feature film converge. In this way, the game itself could be considered groundbreaking. As far as the characters go, we initially decided to scan the actors. We scanned the actors on location in New Orleans with a Los Angeles based company called Eyetronics. Eyetronics utilizes a proprietary technique of shooting the subject with hundreds of digital photographs while projecting a grid on to them through the flash. The scanning process took about 20 minutes per actor. This data enable the artists at Eyetronics to photogrammetrically build virtual counterparts to the actors and the digital photographs are used to create texture maps that are then projected back on to the digital doubles. The interesting result was that the scans looked too much like the actors! We then had to manipulate the digital models to become less real and more “gamey.” Our modeling team led by Raul Moreno had to alter the proportions and facial features to look less like the actors and more like the “idealized” features found in today’s games. We stopped just short of what you see today and opted for a hybrid look which kept the characters a little more grounded in reality. We even altered the textures of the actors by combining them with high resolutiondigital pictures of people on the “Stay Alive” VFX team. All the characters were finished off with CG hair. Digital artist Ralph Knight became our own digital hair stylist, spending two months on hair alone for the hero characters.
Opening Sequence, Lead Artists: Kent Seki [Animator], Brad Friedman [TD] and Tony Lupoi [Compositor].
Hutch's Apartment Sequence, Lead Arists: Duane Floch [Animator], Brad Friedman [TD], and Tony Lupoi [compositor].
What was the approach to make it feel scary?
Ultimately, you take your cue from the director. For “Stay Alive,” Director Brent Bell and co-writer/producer Matt Peterman came in with a list of both movies and games that they found influential and interesting. As avid gamers, Brent and Matt proved to be a great litmus test for the success and believability of the game shots. Many times, we would want to go more cinematic with our shots, but the filmmakers kept us grounded in the gaming world. In addition, Brent and Matt came to the table with a definite idea of where they wanted the picture to go in terms of the effects and pacing. Toward the second half of the post production process, Producer McG added his energy and creativity to the mix for the final push to finish the shots.Visual effects are but one component of creating a mood. Our work is dependent on the editing, music, sound design and performances. We did, however, have moments to influence the horror aspect of the show. At one point, Producer McG gave us the edict to make some porcelain dolls more scary saying, “Find the most F%^*d up guy you know. Set him on fire, and put him on this!” We got the message and had 3D artist Mahito Mizobuchi remodel all of these dolls using his imagination to create a rather disturbing set of dolls.
Scary Dolls Design, Lead Artist: Mahito Mizobuchi
What were the challenging aspects of this production for PLF more well known for it's Previz work?
The single most challenging aspect of “Stay Alive” was creating a compelling narrativefrom the full-screen game shots. “Stay Alive” is based on a yet-to-be-created “next generation” computer game. The game footage had to not only hold up on its own, but more importantly, it had to propel the story forward, often becoming a key storytelling part of the picture. These weren’t just throw away game shots to get us to the next live-action sequence. Rather, the footage offered key clues to the mystery of Elizabeth Bathory, and the believability of the game had to be balanced with cinematic concerns in terms of keeping the audience interested even when no “real” characters were on screen. The first person perspective shots needed to be exciting while the actors’ virtual counterparts had to hold narrative weight. The biggest test came in the third act when Swink (Frankie Muniz) guides Hutch (Jon Foster) via cell phone through Malkus’ House while using the game as an exploration tool. The goal of this sequence is toseamlessly switch between game mode and live-action while keeping the pacing fast and interesting.
Working within a limited budget becomes a challenge in it of itself. There are always things that you want to improve, but you simply run out of money and time. We always try to go the extra mile for each job. For example, the production simply did not have the budget built in for facial animation. We recognized fairly early on that any next generation game would, of course, have facial animation. Animators like Daniel Loeb, Daniel Robichaud, and Brad Friedman took the challenge up themselves to add what we knew was a necessary feature. In the original description of the City of Dead Ghost Girl Chase, the budget alloted for five to ten girls. Duane Floch took that sequence and animated 37 different girls to really take those shots to the next level. Games today also feature a good amount of motion capture. Once again, the budget was not able to facilitate a mocap session. Marko Vukovic constructed a re targeting and blending system to take existing mocap that we had and blend that with key-frame animation. The team consistently impressed me with their ability to creatively address
budget and time constraints.
Opening Sequence, Lead Artists: Marko Vukovich [animator], Brad Friedman [TD], and Kent Seki [compositor].
Hutch's Apartment Sequence, Lead Artists: Duane Floch [Animator] Brad Friedman [TD], and Tony Lupoi [Compositor].
 Tanissa Potrovitza, VFX Producer
Tell us about your involvement in the project and what your responsibilities were on Stay Alive.
I was Pixel Liberation Front’s VFX producer. We were the primary vendor handling about 140 shots, under the direction of Gayle Busby the studio’s VFX producer for the entire movie.Internally at PLF I was responsible for client interactions, managing budgets, scheduling meetings, and coordinating artist feedback. I worked directly with editorial and Pac Title in receiving shot counts and live-action plates. I handled the delivery of the shots both to editorial, as well as the final delivery of Cineons to EFilm. I scheduled DI and film out sessions and generally tried to assist the creative team with what ever they needed. Luckily I had a wonderful production assistant, Beverly Abbott, who helped me out a lot with the nitty gritty of it all.
Tell us about what it involves to handle an entire production as VFX producer.
Of course each project is different, but in an ideal situation, a VFX producer would be involved from the pre-production of a project. From the get-go, they could work on breaking down the script and evaluating needs – what types of effects, cost estimates, timelines, etc. Once underway, the producer could remain on-set, or at least in close contact in order to make sure the VFX team gets what it needs, and if script or shooting changes occur, they are still within budget & available recourses. Otherwise the VFX producer would revise the budget and/or solicit additional vendors. Finally, once in post, the VFX producer oversees the project, while acting as the gateway between the studio/director and the artists. This includes getting feedback on shots, reconciling updated counts lists from editorial, and reorganizing personnel as effect requirements change from shot to shot. Depending on the production, many of these duties are split between a VFX supervisor and VFX producers. In the case of Stay Alive, PLF’s creative director Kent Seki was the VFX supervisor. He traveled to the sets, oversaw VFX shoots that needed green screen elements, tracking markers, etc., and shot tons of digital photos for our artists’ reference, which really helped later on when we had to design virtual sets and objects. Kent also personally jumped in to help many times – from meetings with the studio to final compositing – it was my first big project and he was an enormous help… thanks Kent!
Tell us about organizing film out and screenings feedback.
I found the film outs to be very productive because they let us see our shots as they would be seen in the theater. This confirmed that we were going to have enough latitude for the DI artist to make adjustments to the shots without losing detail. We could also diagnose problems that might be less visible back at the office where we were limited to a standard video projector. It was also a great moral boost to see the shots on film; we would always leave the screening thinking, “Wow, that looks cool!”
What were the most challenging tasks you had to master?
In general, there is a lot of pressure to keep things on schedule and on budget. It’s also very important to keep both the client and artists happy, which can be tricky at times. But I guess my most difficult task to do was to tell an artist that their shot had been cut from the movie. Everyone worked really long and hard on even the shortest of shots, so it sucked to have to be the bearer of bad news.
Miller's Office Sequence, Lead Artists: Daniel Loeb [Animator], Derek Fisher [previz artist], and Kent Seki [compositor]
Opening Sequence, Lead Artists: Marko Vukovicc [Animator], Brad Friedman [TD] and Kent Seki [Compositor]
 Duane Floch, Animation Supervisor
Tell us about your involvement in the project and what your responsibilities were on Stay Alive
I was brought in as the animation supervisor on this project. I had a great team of 6-7 animators, and somewhere around 120 shots involving character animation to supervise. As for my responsibilities, any shot requiring character animation went through me first in the approval process. In addition to supervising the team I also had about 20 shots of my own- mostly involving creepy little girls chasing main characters. I'd discuss each of the shots with Kent Seki, our vfx supervisor, and together we would determine what needed further attention and what was ready to submit to the client. There was a great deal of collaboration between Kent and I in determining the style and feel of not only the character animation but the in-game camera animation as well. Successfully communicating those concepts to the team was very much a part of my job.
What animation choices were made to achieve a game like look for the animation?
The concept was to create a next generation gaming style. While much of that had to do with the look and feel of the environments and characters, we were determined to add as much detail to the character animation as possible in the time allotted. Where camera animation was concerned, we focused on the 1st and 3rd person perspective shots to give them a realistic game feel - if it didn't feel like you were actually playing the game, it was tweaked until it did.
What were the most challenging tasks you had to master?
I would say managing the 3 long shots that involved this gorgeous cemetery environment, two main characters and 37 of those creepy little girls!. There were a couple of other shots involving the girls in which I had to find that look you see alot of in horror films like The Grudge, that disturbing 12fps backwards walk in reverse thing.
Which features of XSI did you find very useful?
I love the animation mixer. It has saved me on any number of occasions - most specifically when I had to transfer a lot of complex animation to an updated rig half way through production.. Using the mixer we were able to get a lot of across the board consistency for walk and run cycles I'm also glad to see that reference models are coming into there own in terms of reliability!
Which areas should be improved?
While I really like working in the fcurve editor, it is aggravating when you add a key to an existing curve and the surrounding key tangents are affected, even if they are locked! That's about the last thing you want to see happen after spending hours adjusting your animation.
 Andy Jones, Chief Technology Officer
Tell us about your involvement in the project and what your responsibilities were on Stay Alive
My current title at PLF is “Chief Technology Officer.” Basically, it's the same job I had with my family, growing up, except now instead of an older sister who needs help printing last night's homework assignment, I have a building full of artists who need help printing shots for a feature film. On Stay Alive, that involved everything from choosing hardware for our computing infrastructure, to revamping our rendering pipeline, to overseeing technical aspects of how we model, animate, render and
composite. I also spent an embarrassingly long time doing some fluid simulation work. I suppose you could say that if PLF were the Ghostbusters, I would be an Egon Spengler. I say “an” Egon Spengler because Brad Friedman would also be Egon. And Marko Vukovic would be some sort of Peter Venkman/Egon Spengler hybrid. Honestly, I think my single biggest contribution to Stay Alive was finding Brad spouting some bit of Egon-worthy gibberish on the XSI mailing list and telling Kent to call him and try to lure him away from his last semester at NYU. It was really great having both him and Marko around, especially when a bunch of our character rigs somehow became possessed by the Keymaster.
Which custom tools did you develop to improve the production process?
Going into Stay Alive, we were already using some custom tools I'd made for previous projects. Most notably, the backbone of our rendering strategy is a shader I wrote to allow us to control our pass setups with expressions instead of the slower, clunkier and less flexible material overrides applied to partitions. Technically, you can do the same thing without a custom shader, which is what we did on “Sky Captain and The World of Tomorrow,” but it becomes a lot more difficult for the artist to do anything creative in the render tree without breaking it.
Due to the nature of this particular project (i.e., non-photo-realistic effects on a fairly limited budget), we focused most of our development time on revamping our existing pipeline to be as non-linear, automated and reliable as possible. In general, the less time you have budgeted for each shot, the more important it becomes to automate repetitive tasks, leaving as much time as possible for the creative stuff like animating, lighting and compositing. The bulk of the development we did that ended up getting used in this job was for some tools that enable us to automatically sort our assets into a full-blown pass setup, with multiple partitions per pass. For people who don't render with a lot of passes, this might not seem like a big deal, but it's not uncommon for us to have around 10 passes per layer of a shot, each with around 10 partitions. I'll skip the details, as Brad already gives them in his answer to this question. There were a couple of other tools I made directly in response to Stay Alive, that still haven't really hit our production floor yet. That's sort of how it goes with a lot of development, since there can be a certain degree of risk and/or overhead associated with implementing new technology in the middle of a show. The big example that comes to mind is a shader I made for rendering hair with framebuffers. For those who don't know what framebuffers are, they're basically a way to make mental ray render multiple passes (in the XSI sense) simultaneously. Pass Channels in XSI 5 are essentially just framebuffers wrapped up in a Happy Meal for mass consumption. Anyway, after doing a job with hair, we realized that it's highly desirable to break apart the hair render into separate passes that can be combined in a 2D application. We've actually found that to be true of pretty much everything (hence the 10 passes per shot layer), but in the case of hair, rendering additional passes without framebuffers is especially costly, since generating the hair geometry can sometimes take as long as the render itself.
What were the most challenging tasks you had to master?
Fluid simulation, but I wouldn't really say I “mastered” it. “Mustered” is more like it. The handful of times I've had to do fluid simulation, it's often been for close-up shots, which, in my opinion, are still really hard to do well with simulation, even with great tools like RealFlow. No one ever asks me for the two spouts of blue and yellow liquid pouring into a glass and mixing around, like in the RealFlow tutorials. Instead, they always seem to want something that falls slightly out of the realm of what the existing tools are good at. In this case, the problem was that the shot started with a view of a deep pool of blood and ended with a close-up of blood flowing through the nooks of a small metal crest as it emerged from the pool. This is a tricky thing to do, because simulating a pool of blood with enough precision to give proper flow around a detailed object would have been prohibitively slow for us. The strategy I ended up using was to render a depth matte of a mesh from a high-resolution RealFlow simulation for just the close-up particles, combine it with a depth matte for the larger body of liquid in 2D, then use the result as a displacement map for a liquid surface in the final render. It worked pretty well for the special case I was dealing with, where you don't ever need to see the underside of the liquid. A bonus of this approach was that I could manually roto in changes to the simulation at the 2D phase, which was easier than trying to do modeling on a RealFlow mesh -- especially since I'm just about the worst modeler on the planet.
Which features of XSI did you find useful for your tasks?
Ever since I discovered that it existed, I've been a fan of the FXTree, and it proved very useful for the little fluid simulation displacement technique I described above. One of my other favorite XSI features is the little connection icon that lets you drive parameters with sources like weight maps, and texture maps. A little rain cloud appears over my head whenever I don't see it next to a parameter. When combined with the FXTree, you can do some really amazing stuff with those little icons. For example, at one point I did a really cool particle test, where I pulled a difference key on adjacent frames of a sequence of plates of some rose petals burning and used it to emit smoke particles along the burning edge of the petals. There's also a scene in the movie where Swink, Frankie Muniz's character, is customizing his game character, in which Ralf Knight used a push deformer with a weightmap to dynamically animate the buff-ness of the character mesh. The effect was pretty much exactly what we needed, as it quickly starts to look ridiculous as Swink tries to make his character unreasonably strong.
I also have to mention BatchServe. I know some people have gotten burned by it in the past, but we've used it to render several shows now and it works pretty much flawlessly. There's a big disclaimer though, which is that I've had to make significant changes to the code to work around some bugs. The relevant code is just vbscript and php, though, so making changes isn't really such a big deal if you know what you're doing. Plus, while you're at it, you can add features. For example, in our customized render control script, we save huge amounts of networking overhead by syncing all our texture files to the local drive of the render machine. There are so many things XSI does right. I'm mentioning these because a lot of people don't use them as much as I do, but really I could go on for pages.
Which areas of the software should be improved?
The FXTree is a really powerful tool, but it could be about twice as powerful and five times as useful with some additional development. I've whined about it for a while now, but we could really use a text tool in there for doing reticles. And the floating point support is pretty much non-existent in the current set of nodes. You can get HDR and OpenEXR images in, but that's about it. Even AfterEffects supports floating point compositing now, so it must be time to jump on the bandwagon.
Floating point support in the FXTree is just one part of a whole set of changes I'd like to see made to provide XSI with a consistent linear rendering workflow. People always say, “CG is linear,” and they're right, since most of the shading methods we all use (lambert, phong, etc.) are intended to represent idealized linear light. However, people are still picking colors and painting textures in PC gamma space, so the resulting renders are an intertwined mess of linear light data and PC gamma color data. What's missing from XSI, as I see it, is a convenient way to get color data into XSI in linear form, and a way to quickly view rendered results with gamma correction for a PC monitor, similar to the view lut capability available in most compositing packages. We've actually been in the process of developing most of the tools we need ourselves, and we're using them on our current project, but ultimately, there are things we'd like to do that will never be as elegant as they could be if Softimage implemented them directly.
Lastly, it should be possible to put clusters into partitions. I've seen countless tutorials on crazy methods to deal with avoiding cluster materials because they make pass setups difficult. Solving this problem is like trying to design a car without any wheels. There are ways to make it work, and I'll use them if I have to, but can we please just have some wheels already?
Which direction should the software take in future developments?
For the most part, I'm pretty happy with the direction XSI is going in. I think the special projects guys, in particular, have done some really amazing stuff lately, with Face Robot. I attended a motion capture summit this fall with an assortment of really smart, knowledgeable people in the effects industry, and probably the biggest thing everyone unanimously agreed on was that facial motion capture is a really big unsolved problem in visual effects. I was very pleased to see Softimage stepping into the lead in this area.
As far as XSI goes, I'd really like to see Softimage take some time and re-evaluate their tools in the context of actual production pipelines to try to absorb some of the functionality that currently exists only as customizations and workarounds. Right now, a pipeline based exclusively around native XSI is a bit of a roller-coaster ride, with some really high peaks, but also some pretty low valleys and occasional broken rails. And even with customization, there are still some pretty low moments. Reference models and shape animation, for example. Personally, I'd rather see those problem areas ironed out than see more features added. I understand, though, that shiny new features are often what sells software, and what keeps users like me drooling about new releases.
 Marko Vukovic, Technical Director
Tell us about your involvement in the project and what your responsibilities were on Stay Alive
I had worked with Kent Seki in New York where we met on a project at Nickelodeon and then later we worked together at Psyop. I really enjoyed working with him and we had a good working relationship on previous projects. I mentioned to him a few months back that I was looking to get involved in feature work and he said that he had a VFX horror film in the works. It sounded very interesting. Plus I loved the idea of working on a CG horror flick!!! My role on the project was to try and make the mocap and animation pipeline a happy harmony, however I ended up being involved in many different aspects of the production including vfx supervising on set in New Orleans.
Tell us about the rigging process for the characters and what custom tools you developed
We basically used the tools already provided by softimage! Isner's biped rigs are great at getting setup quickly. I used the rig before on many projects and I was very familiar with the stengths and weaknesses. We slightly modified the rig, added some extra controls that the animators wanted, but for the most part the rig was left unchanged. I built a custom synoptic view for both the characters and the POV camera for in-game gameplay. I also wrote a bunch of scripts to make the rigging process faster and standardised. Naming convention was very important when it comes to automation, it just makes everything that much more efficient.
Tell us about how you used rigid bodies to create a ragdoll
This was fun. I started playing around with rigid bodies to test how stable they were and if it was possible to use them for character animation. At first I wasn't planning on using it on the characters but I quickly realized that I could hook up isners rig to RBD and have the simulation drive the character. There was one shot in particlular in the opening sequence where Loomis character gets thrown over the banister and is hung by his neck.... the rigid bodies proved to be a very efficient way to generate a performance that would have been too difficult / time consuming to generate. It was very easy to get several performances out of the RBD just be tweaking gravity or mass. I ended up blending two different simulations using the animation mixer and added some additional keyframe animation to fix certain intersecting issues.... overall the RBD got me 80 % there without ANY animation!!!
By plotting the RBD I was able to edit the anim curves using the animation mixer, a fantastic tool for re-timing and cutting together different animation clips. Who said you can't direct dynamic simulations???
Talk about using Realflow to create blood like fluids
Real slow.... honestly to be fair, real flow is a fantastic liquids simulator. One day when machines become faster it might even be realistic to use it in production. We got good results in the end but it was a VERY slow process. It would usually take a day to just simulate the particles, it looked great after it was done but then if you wanted to tweak some setting you have to wait another day for it to process... not ideal but it works.. Setting up several version overnight became the standard.
What were the most challenging tasks you had to master?
Dealing with mocap data. The tricky thing about mocap is that the data is never accurate to human anatomy. The sensors that the performers wear live above the skin so you have to compensate for that offset on your 3d model. Plus those suits are totally goofy looking. Once the mocap was working the next task, and this was Kent's main concern, was to be able to blend mocap and animation data. This proved to be a tricky task. We needed a way to have the characters driven by mocap data but also have the flexibility to animate on top of the mocap. The animation mixer played an important role here. I could speed up or slow down animation, disable certain joints or even animate on top of mocap!!! The rig I built for the characters allowed us to bake out the mocap with the animation on top. It was basically a retargeting and offset rig. In the opening sequence where loomis goes over the banister I blended several different animations and RBD simulations together to get the action, once this primary animation was in place Brad helped to setup the cloth simulation.
Which features of XSI did you find very useful?
Scripting, scripting and scripting!!!! Envelope painting, smoothing and weight editing, rig standards for biped characters. Also relaxing the mesh was a huge time saver for texture mapping.
Which areas should be improved?
Ascii format of scene files would be the best feature for 6.0... Retargeting tools which are currently missing in XSI for mocap, reference models, reference materials, refrence preferences and any setting in XSI should have the ability to be overriden or inherited... this way a TD could setup scene settings that everyone else in the pipeline inherits including geometry, materials, preferences etc. More buffers for the render region would be a nice addition.
 Brad Friedman, Technical Director
Tell us about your involvement in the project and what your responsibilities were on Stay Alive
Originally, I was hired on with the hope that I would be able to help out with the rendering process on Stay Alive. However, by the end of the production, I was serving as a technical director across the character, finalling and rendering areas of production. I also spent some time modeling and texturing. For this job, I had to switch hit between being a specialist TD and being a generalist, on a daily basis.
What were the most challenging tasks you had to master?
Well for starters, getting my footing in a professional environment was very challenging. It was jarring to be thrown from an educational environment into production. Figuring out how to deal with my superiors and colleagues was a daunting task. There I was, fresh out of school. There they were, veterans. FILM VETERANS! Who was I to work with them? Who was I to tell them anything at all? In school I was used to being the 'fix it guy.' But professionally, I had no idea where I stood. Also, working my way into someone else's idea of how to render was confusing at first. The pass system at PLF, is very different from how most people are taught to use mental ray. I had to wrap my head around what they were doing and why, before I could really work with it and start to innovate with it.
What custom tools did you develop to improve the production process?
Andy Jones and I have put our heads together on quite a lot of tools and ideas since I've been at PLF. I'm sure his responses to similar questions will overlap my own. Generally, Andy and I play a lot of devils advocate for one another. When we both agree on something, Kent will do us the favor of playing devils advocate for both of us. To the outsider, I'll bet we look like we're about to explode into an all out brawl on a regular basis. However, when we're done, we usually have a thoroughly vetted design concept thats better for having gone through the process. Then, Andy codes the hard parts and I take the credit. Its a perfect symbiotic relationship. I guess the first major pipeline development I contributed to PLF was very simple, and yet in retrospect, its rather profound. I suggested we use some open source software called Subversion (http://subversion.tigris.org/) to manage our XSI workgroup. That way, the workgroup would live locally but be version-able. This allows computers to go off
site easily and avoids the nightmare of the corrupt network bound workgroup, that takes everyone down simultaneously. Ever since that day, we've been using it. Andy and I can develop directly into our local workgroup and test thoroughly before committing our changes to the rest of the production. Deployment is almost automatic. I can't imagine how it would be to work without it at this point. In the middle of the production process on Stay Alive it became rather apparent that the pass system we were using, while powerful, was also a burden to maintain and implement across the production. Simply adding TDs to the render pipeline was not going to make it go anywhere near as fast as it needed to be. Therefore, I drew up a design based on some techniques I had been using in the development of my custom rigging tools. The design was for a more automated system, based on the existing one. Without going into too much detail: our system depends on sorting the contents of our assets (models) by their material properties, such that an automatic system can then build passes just before we render our scenes. This involves encoding a substantial amount of meta data into the XSI scenes in a way that artists can understand and work with. By doing this, we pushed a huge amount of the work of setting up passes, back to the artists responsible for creating the assets. The actual development of this system involved a rather large amount of custom tool coding in Python. Andy and I tag-teamed the development and implementation of the system.
In my role as the main Character TD for the latter half of show, I developed a lot of little tools here and there to assist with complex character interactions and rendering solutions. None of them stand out as particularly worthy of note. Its just part of any TD's daily workload.
Tell us about creating all different hair styles for the characters
While I'm not directly responsible for creating the hair styles, I was heavily involved with the hair process for the show. We had a great artist, Ralf Knight, who did nothing but style hair for what seemed like months on end. He did a great job of coaxing the hairs into place with a combination of hair sculpting tools and modeling tools (while the hair was in stretchy mode). All of this was done in XSI 4.2, which had a more primitive hair sculpting tool set than 5.0 does. I honestly don't know how he stayed sane. Hair rendering in the show was done with a combination of the standard geometry hair shader in XSI, and some muhHair shading for select purposes. It was painful however. Hair rendering, in general, is a painful process. The balance of look, with processing power, and 32bit process memory limits, was enough to make me doubtful on occasion. But we made it work and I think it looks good.
Which features of XSI did you find very useful?
What's XSI? We made this movie in Microsoft paint. I like the paint bucket and the brush tool the best. I would be remiss if I didn't mention Syflex. We used Syflex exclusively for our cloth simulation on the show and its inclusion in XSI Advanced was a significant boon to our production. I am a big fan of the relax deformer. Seriously... the best deformer I've ever used. I can't begin to list all the things it has done for me on a regular basis. Reference models are amazing. Our entire pipeline is now based on reference models. It was a serious leap of faith to trust that they'd work for us. And in fact, they often didn't work. We had to work around a lot of problems and bugs.
We had to change a lot of standards and practices. However, the overall benefit did outweigh the extra frustration and lost time. Especially, when we got to the rendering portion of the show. We used the XSI pass system extensively, though not in the way it really was intended. It says a lot that the system was flexible enough to allow us to build our own system on top. Python was our scripting technology of choice. Its not perfect but it did preform admirably.
Which areas should be improved?
I'm no stranger to the XSI mailing list and I don't shy away from criticism of the software. So, a lot of what I have to say is already known. For example, Andy and I evangelized a point of view of linear rendering workflow recently on the list and I feel strongly that XSI needs to transcend the issue and make a linear rendering pipe the standard, default way of working in XSI. All the gamma issues should be put out in the front of the UI rather than having them ignored and buried in the options and nodes. It's important and the current state of affairs should not be considered “good enough” as they are.
The Animation Mixer needs improvement IMHO. I think its light years beyond the other 3D programs out there. However, that doesn't mean its OK to let it linger. Nested compound clips need full support rather than the odd tacit support they currently have. Animating and editing FCurves within a clip should be just as easy as editing FCurves on the timeline, with full interactivity with the viewport and current object selection. Time should be spent making sure all clips of all types can easily be imported and exported from the Mixer (they currently don't all work). More effort should be put into animation filtering with the Mixer. Adding a layer of noise to a clip should be as simple as a few button clicks in a nice UI for your average animator.
Shape animation, and how it works in a reference model world, needs to be examined and the problems need to be solved. Both animators and character TDs need to be able to author shapes and work with them back and forth and in parallel. Currently, this is very difficult to accomplish with reference models.
Abigail's Van Sequence, Lead Artist: Mike "Berzerkle" Werckle [Animator] , Kent Seki [3D Compositor] and Diana Hinek [2D Compositor].
Cemetery Sequence, Lead Artist: Raul Moreno [Compositor], Casey Hunt [Matte Painter], and Reza Rasoli [Tracking].
 Raul Moreno, 2D Compositing Supervisor / 3D Artist
Tell us about your involvement in the project and what your responsibilities were on Stay Alive
Initially I was responsible for the look of the characters for our "game". After this phase was finished I switched roles to the lead 2D compositor, finishing the live action effects shots. As Kent said earlier the characters looked to much like actors in the film, so I had to alter them subtly. We took digital photos of some of the people on the team and integrated them into the characters and modified the geometry. Needless to say there were many revisions for the characters, until we got the look we wanted.
What were the most challenging tasks you had to master?
The Cemetery Sequence immediately comes to mind. A long tracking shot (1154 frames) that needed to be composited over a really nice matte painting that Casey Hunt did. It tracked very nicely until the actors walk in front of and push open an old wrought iron gate with ivy and many of other things hanging in from it, really making it impossible after this point to use any type of automated tracking software. After they walk through this gate we had to hand track. This meant about 450 frames had to be tracked by hand and there was enough eccentric movement in the camera to make it real tricky. It was very tough to lock the matte painting down. Reza Rasoli did most of the tracking. Reza got track very close and I picked up the shot after that.We spent much time on this shot, from Claire Lin having to Roto all the ivy, wrought iron, actors, etc. from the shot, Casey's matte painting, tracking, color correction and integration. It was by far the toughest shot for me on this project.
Which features of XSI did you find very useful?
I really like the modeler in XSI. It's a joy, easy to use yet very powerful. I teach a class at Otis College here in Los Angeles where we focus on digital sets and environments using XSI. The students are able to produce some complex models in a very short time.
Which areas should be improved?
The Texture/UV Editor could use some improvements and updates. A pelt tool would be a good start. The ability to change units. Feet w/ fractional inches, Feet w/ decimal inches, meters, etc.
The ability to change grid spacing with major lines every Nth grid line and change the color of them. A built in measuring tool. It would be nice to see the compositing environment develop a bit more. If you are familiar with Shake it takes very little time to pick up.
 Tamir Diab, 3D Artist
Tell us about your involvement in the project and what your responsibilities were on Stay Alive
I was hired by PLF as a lighter but with more generalist responsibilties such as modeling, texturing, and some layout.
What were the most challenging tasks you had to master?
The most challenging tasks I had to master would have to be the detail within the modeling phase. We had to model assets for the film from set and reference photos taken by Mr. Kent Seki and the PLF on-site crew. I needed to model a horse carriage that would closely resemble the one in the live plates. Details like rivets, and how metallic surfaces really work was fun to get into. I also needed to model the carriage in a way that would best the rigger and animator's needs. In other words, it had to look right but also be functional. I have to say that the rigger Marko Vukovic did an amazing job. Since I had been out of the modeling arena for some time, it was great to get input and share tricks from other PLF staff.
Which features of XSI did you find very useful?
I found the XSI modeling tools very useful and intuitive. The Carver tool written by Ralf Knight helped make the modeling process much easier. Especially while modeling symmetrical objects, thes tools literally cut time in half. Also, Marko Vukovic had a real handy extrude along curve tool that helped me with all the metal rods and window frames on the carriage. It was also extremely helpful while modeling chandeliers that were to be used in a home's interior.
 Keely Colcleugh, 3D Artist
Tell us about your involvement in the project and what your responsibilities were on Stay Alive
This was my first ever project as a CG artist! I was involved in Stay Alive from May 2005 to September 2005 . My primary responsibilites included the creation of assets and environments..more specifically making models of buildings and props then texturing them.
What were the most challenging tasks you had to master?
As an entry level artist just about everything was a challenge. Working with a strong team of senior artists and producers with years of experience (and tons of patience) behind them helped immensely.
Which features of XSI did you find very useful?
Coming from a more traditional modeling background and relying heavily on orthographic projections (plan, section, elevation drawings) to describe 3D shapes, I've been amazed by how quickly XSI can bring a drawing, storyboard or idea to life. In virtually no time we were able to try out configurations of huge environments, play with props arrangements and update eveything using reference models in the hero scenes.
Which areas should be improved?
The texture editor seemed to have a few minor glitches and I'm always a bit frustrated with the texture path problem. However PLf's well-organized pipeline and culture of communication between all teams kept slow-downs and problems to a minimum.
 Nick Loizides, 3D Artist
Tell us about your involvement in the project and what your responsibilities were on Stay Alive.
I was involved in the asset creation of environments, props, texturing and lighting that were used for the CG game in the movie.
When I arrived at PLF, Stay Alive had just started filming. My first task was to model environments that I then textured and lit and FTPed them to Kent on location in New Orleans. The idea was to show the Director and production designer game concepts of the interiors and exteriors.
I then moved onto modeling, texturing additional sets and adding more details to the approved concept environments. And finally, I did some preliminary lighting.
Tell us about modeling entire interiors and dealing with large amount of objects and big scenes.
Although this was a game played in the movie, the film makers wanted a better than “nex gen” look for the game. The great thing about this was I could put a lot more detail into the models. Things like, beveling everything, warped and bent wood and detailed chains hanging everywhere. A lot of objects I created by projecting the photo reference image onto a grid and then modeled and extruded off of that. This cut down the time to UV as I created geometry based on the image. I’m very happy with the results of this method.
Modeling and texturing was made easier for me because I had good photo references and images from the art department. It was a great challenge to study a photo of and interior, furniture, accessories, and recreate it in 3D. Of course, the scenes became really big and hard to move around. I made groups and layers of the room contents and turn them on or off when needed.
What were the most challenging tasks you had to master?
I would say the pass-mixer, which is a PLF propriety shader system. It’s big and at first over whelming, but very useful when achieving a good render. Another challenge was lighting big scenes. We had so much stuff in the scene especially the exteriors with all the trees. I had to apply a lambert material to everything and light the scene that way.
Which features of XSI did you find very useful?
I’m relatively new to XSI so my choices might seem a little simple. But, over other apps, the texture editor. It’s very easy to see where everything is and what it’s doing. Also, my favorite, the split edge loop tool, which was only just introduced to Maya 7, but XSI has had it for a long time. Just select the ] key and middle mouse click on an edge. I would be lost with out this cool tool.
In lighting, I like that you can see and tweak the light cone and fall off without going to the controls. Select the light and push the b key.
Which areas should be improved?
The ambient default setting is too high. I always set it to zero. And, in the lighting controls, when you turn on shadows, the umbra default is set very high. I think it should be set lower and if you want to see into the shadows you can turn it up. I’d like to see a customizable right click menu. For example, if you’re modeling you could pick your most used tools.
When you use two monitors and have the Explorer – Render Tree open on the second monitor, have the render tree refresh automatically when you select an object. I always have to click refresh otherwise I inadvertently edit the wrong set of shaders.
Of course there might be ways to change all of these things, I just might not have found them yet…
Credits
Kent Seki - VFX Supervisor
VISUAL EFFECTS BY PIXEL LIBERATION FRONT
Sean Cushing - Executive Producer
Tanissa Potrovitza - VFX Producer
TECHNICAL DIRECTORS
Brad Friedman
Andy Jones
Marko Vukovic
3D ARTISTS
Glenn Burton
Keely Colcleugh
Tamir Diab
Mahito Mizobuchi
Ralph Knight
Nick Loizides
Scott McGinley
Michael Meyers
Eric Schoengarth
Raffaele Scaduto-Mendola
ANIMATORS
Duane Floch - Animation Supervisor
Derek Fisher
Daniel Loeb
Daniel Robichaud
Mike Neely
Mike Werckle
COMPOSITORS
Raul Moreno - 2D Compositing Supervisor
Tony Lupoi - 3D Compositing Supervisor
Brian Battles
Diana Hinek
Shant Jordan
Wenshan "Claire" Lin
Jeff Olm
James Payfer
Freddie Vaziri
PRODUCTION STAFF
Beverly Abbott - Production Assistant
Sim Poo Lem - IT Assistant
Abby White - Office Manager
INTERNS
Casey Hunt
Gary Kavanagh
Reza Rasoli
Related Links
Pixel Liberation Front
Stay Alive Trailer
Visual Effects clips
Discuss Article at the forum
| |
|
