Trip the Light Fantastic - Linux in the Special Effects Industry
In early 1998, Linux users began to be aware that their operating system of choice was making a serious impact in the field of movie production. Daryll Strauss had written an article for Linux Journal, in which he described how Digital Domain had rendered scenes for the box-office busting Titanic on a farm of Linux boxes. Eight years later, the foothold of Linux in the industry is well established, and so we paid a visit to this pioneering studio to talk to Doug Roble, Creative Director of Software, and Dave Fallon, manager of the Systems Administration department at Hollywood effects studio Digital Domain, to find out how it happened.
Daniel James: Doug - what year did you join Digital Domain?
Doug: I joined the company in 1993. Because of turnover there are only two people in the whole company who have been here longer than I have - the President, and one of the Vice Presidents. True Lies was our first very large film. Before Titanic we had also done Apollo 13 and Interview with a Vampire. James Cameron had gone to a Polish shipyard and asked them how much it would cost to retro-fit an old tanker so that it looked like the Titanic, at least on the outside, so that he could sail it around the ocean. It would have cost a lot of money, but it wouldn't have given him the kind of control that he had.
DJ: So it was costed as a realistic option then?
Doug: Yes, to rebuild the Titanic for that movie. Remember at the time, that movie was the most expensive film ever made. I think in total dollars, if you count for inflation, Cleopatra was the most expensive film ever made.
DJ: But it was worth it, wasn't it?
Doug: $1.8 billion dollars world wide - there is nothing even close to it. I did read in Variety they were speculating that King Kong might reach Titanic heights, but it's a little early. King Kong is cool, but he is no Leo - you have to get the teenage girls going to see it again and again.
DJ: Before CGI, films relied on models and those kind of techniques, didn't they? That often wasn't entirely convincing to the cinema goer was it? People would say: Ha! That's a model!
Doug: Even now Digital Domain is a full feature effects place, we still build models. Models at times are very cost effective, and it's very easy - the model shop guys are stunningly amazing. You will see them just go off and build and aeroplane and fly it around and film it. It's very difficult for CGI to compete with them even now. If you go out across the street [from Digital Domain's headquarters in Venice, California] you will see a couple of aeroplanes that we build for Stealth, and these models were used quite a bit to blow up, or to do effects that are still tricky for the computer to do.
Dave: The great part is the combination of the two. With special effects you can clean the little bits of the model that made you go 'oh that's a model, that doesn't make any sense.' Putting them all together really makes it work.
Doug: I just watched one of our old films that came out the same time as Titanic - Fifth Element - and in that there is a lot of combination model work and CGI work. A great example of that is the New York taxi chase; the buildings are models for the most part, some of them are digital. But they added so much stuff to it; all the cars were models or digital cars. Just layer things on top of everything and you make what looks like the cheesy little model look like the real thing; the effects still hold up.
DJ: Is it overstating the case to say Titanic was a breakthrough film, technically?
Doug: Remember Titanic was a model - the only time you saw a digital boat in Titanic was when for the simulation of the sinking (they showed it at the very beginning) the boat broke in half. The people who are searching for the Titanic had a little video display where they explained what happened; it sort of set the stage for the story. Any time you saw Leo or Kate walking along on the top of the Titanic, that is them on a blue screen being married to a model, and the digital parts of it were the water, the people on the deck, the smoke - all the environment that made it look real. The Titanic itself was a model for the most part.
DJ: Was that scaled down?
Doug: It was scaled down, but it was still monstrous. We had a sound stage over on Playa Del Ray at the old Hughes Airport, in a huge hangar - the thing was 40 feet long. We had this huge model and these accurate camera rigs; they are super accurate, you could move them across an entire football field, back and forth, and they will return to their starting position to a couple of millimetres, because they laser calibrate them. It's very neat - it has to be, otherwise you will see stuff jiggling on the screen. So they set this up; they had the model, they had this accurate model mover. They filmed a pass of the Titanic model in the morning, they backed everything up, changed the lighting and filmed it in the exact same pass. You never do something just once - you film it once with a certain kind of lighting, then you composit it to make it look real.
And they noticed that every time they did this, they weren't lining up. They recallibrated the model ray mover over and over again, and something wasn't working. You would move it once, do it again, and over in our stages it would work perfectly. Over there, finally they realised that Playa Del Ray was right next to the ocean, and it's wet; so when the tide went out the cement floor sank a bit, and when the tide came in the cement floor rose up a bit - and you could see it on film. So some of the problems were in live action shooting - but it's a lot of data to build that in the computer.
DJ: Who took the decision to use Linux on the film Titanic?
Doug: That was a combination of a bunch of people. Daryll Strauss was a big proponent of that - he was the software manager at the time. He had done a lot of early work with Linux, and saw that it had a lot of advantages over the SGI operating system that we were working on at the time. We had some people come over from another company as new hires, and they brought their computers with them. They brought just PCs running Windows, and they were cheaper and faster than the SGIs we had purchased. They were running rings around the artists who were stuck using SGIs, until we looked at the cheap fast hardware that you can get now - it was the transition between the supercomputer SGI and the low cost, fast chips. We said 'that's great, but we are still a Unix place - we don't want to transition all the way to Windows. All the internal networks are Unix.' Linux was popping its head up at the time and so we started experimenting with the cheap boxes. Easy operating system - well, not quite easy - it was quite rough, and Daryll had to do some kernel hacking to get it to work. We don't want to do kernel hacking!
DJ: So was that on the artist desktop, or the render farm?
Doug: No, initially it was on the render farm. The user interface and the tools were there, things like Maya and Houdini. Houdini might have been first.
Dave: We started with the Alpha farm - DEC Alpha workstations, that was our first big Linux render farm.
DJ: Dave - at what point did you come into the company?
Dave: I came in three years ago as the Linux system administrator, and ended up being promoted to the manager of the systems department, so now I get to deal with all the other problems as well!
DJ: I remember reading that the first Linux render farm was a mixture of Windows NT and some Alpha boxes, and there was a bit of everything...
Dave: That was before my time, some of the Alphas were Windows, but for the most part the render farm was Linux. We had Windows workstations.
Doug: I can remember some of the Alpha machines. The people that we hired who brought their own machines were all Windows users, and they came into the commercial division. Then the Lightwave application really made an impact on our facility; it was very fast at doing what it did, and I know they used Lightwave to do some of the stuff on Titanic. I remember there was this constant struggle over this big Alpha farm, and how much of it would be Linux and how much Windows. RenderMan ran on Linux, but some of the other things didn't, and I think Lightwave was one of them. It was very exciting, but it was very frustrating. Daryll was a big guy, and he would go stomping through the company when things would be down. It made a huge impact. I don't think that after that we bought many SGI machines.
Dave: Only because we have to - there are a couple of specific software packages that only run on the SGI machines. We bought three Tezros last year because we needed them for these specific applications. There's a similar situation with Avid - we buy a Mac just because it's for Avid - but beyond that, on a workstation level we don't at all any more.
DJ: How many Linux boxes do you have throughout the entire company?
Dave: We have about a thousand machines in the facility; everything is dual-boot, Windows and Linux. So at any given time half of them run Windows and half run Linux.
DJ: Is that down to individual artist preference?
Dave: Sometimes, depending on the artist - usually, it tends to be due to that particular show's preference, whatever they're doing. Some things are better or easier in Linux, and some things are easier - or possible - in Windows, depending what they are. So all the Lightwave artists use Windows, Maya is split about 50/50, but Houdini is the one where people really seem to prefer Linux. On Nuke, people tend to be Linux users for the most part.
DJ: What kind of hardware are they running, typically?
Dave: We do workstation purchases in batches, on a yearly basis. The slow machines trickle down and end with the facility people or a programmer, versus an artist. Our latest are dual Xeon 3GHz machines, 4GB RAM and pretty big hard drives. We use nVidia Quadro video cards.
DJ: I was going to ask you about storage, as presumably you're moving around vast amounts of data the whole time?
Dave: Absolutely, we have around 24 to 26 terabytes of disc space in the facility, and it's all used. We are constantly running out of disc space, and no-one ever wants to go back and do the accounting process of figuring out what needs to be backed up and what we can get rid of. We use NetApp for our main disc servers, and have a few Linux servers for our second-tier storage.
DJ: On those thousand machines, how many artists are there, and how many software engineers?
Doug: The programming that goes on in the company varies. Ten people in the software department are full-time devoted to writing code. Then we have a couple of people we call the engineering pipeline department, who are also writing code; a lot of script based stuff, where as the software guys are C++ jocks. Then we have Technical Directors - these are the people who span the divide between the tech and the artists. They're pretty much artists who know how to program. They get more attached to individual shots or shows, and work on tools that the shows specifically need. While the software department is people who are writing tools for the whole facility. We recognise a need and we try to address it. TD's write plugins for every application that we purchase, or stand-alone tools. So the number of people writing code is probably in the 30's to 40's. The systems department is also writing a lot of code.
DJ: So the majority of people in the facility are on the creative, artistic side?
Dave: Absolutely - they're paying the bills! It ranges a lot, depending on the size of the projects and the amount of work we've got coming in, but up to four hundred, or four hundred and fifty artists probably.
Doug: The company changes size rapidly; we have a core staff list of about two hundred and fifty. But then as projects come in, there is a large pool of artists that roam around LA and California that follow the shows. The biggest we've ever got was during Titanic. We had three monstrous shows for the time; Titanic, Fifth Element and Dante's Peak, and the company got up to about 1200 people on the payroll at once. We're about a mid-sized company - Sony is over a thousand people now I think, ILM is in the 600-700 range. Digital Domain and Rhythm and Hues are about the right size for our privately-held nature. Rhythm and Hues and Digital Domain are kind of unique in that we are small non-production-facility attached shows. ILM has LucasArts, Sony has Sony, Dreamworks has Dreamworks...
DJ: So you're relatively independent?
Doug: Yes, we're very independent - and very dependent on all the films coming our way.
DJ: After the Titanic render farm got well known, how did you find out that the other studios were starting to pick up Linux as well?
Doug: Everybody talks - it's a very tight-knit community, with people moving back and forth from one show to another. We were also very big proponents of Linux; Daryll and everybody here got a lot of press for using Linux, and we made a big splash at SIGGRAPH. Certainly the research community noticed that we were using Linux, because they were using Linux.
Dave: Everyone was running into the same set of problems as well. It wasn't unique to us that SGI machines were more expensive and not as fast, and everyone was coming to the same conclusions. We were an SGI house because that was the only thing you could be at the time, and now we want to keep our UNIX experience and keep some of the abilities, tools and knowledge that we have, but we don't want to switch over to Windows. So here's this Linux platform that we hope will continue growing and maturing - and here we are.
DJ: I spoke to someone at Dreamworks who told me something similar - he said that they used to have an SGI machine and a Mac on every artist desktop, and then they found they could replace both machines with a single Linux machine that was faster than both.
Doug: Cheaper too!
DJ: They re-equip for every movie, every 18 months or so, and the cost of the hardware alone was making the difference, because of this turnover of desktops.
Dave: We have about a three year production cycle for workstations, but they tend to move down. The top-tier artists will have the latest machines every year, then the next group down will get the 'B grade' machines and so on. Then the software and systems guys get the 'D' grade machines!
DJ: Presumably in the past there were technologies that you would rather keep to yourself, for reasons of competitive advantage. How much of a difference did Linux make to that culture?
Doug: We have this big conference called SIGGRAPH every year, and it's a pretty cool place to be, where everybody tries to show off. It's a very show-off industry, and you're showing off the latest effects that you're doing. Everything that you're doing eventually makes it up on the screen; when we go to movies, it's like 'How did they do that? Oh I see.' At SIGGRAPH everybody's talking, and there's a lot of presentations by production companies, and everybody pays attention. So there's always been a sort of implicit sharing of technology, or at least the concepts of technology, in this industry. Also, the problems are hard, so we can say 'We're using fluid dynamics to do the floods in various films that we're working on,' and we can say that pretty confidently because we know it's difficult to put together. A lot of these companies, while they have software developers, they don't have a huge amount - they don't have IBM numbers of developers - so it's difficult to pick up new pieces of technology as they get bigger and bigger.
DJ: So even though the effects houses are relatively large, they still don't have enough developers to do all the problems all by themselves?
Doug: You always pick and choose what you want to try and tackle.
Dave: There are a few things that help share the information. One is SIGGRAPH and showing off, people wanting to do their best, and two is the close nature of things. Invariably, someone working at Digital Domain will go and work at Dreamworks next week. Finally, everyone's moving forward so fast that by the time you try to figure out someone else's ideas and thoughts, they've already moved on to the next thing. So go ahead, have it! - you can spend your time figuring out last week's stuff, while we work out next week's stuff. Basic infrastructure problems that people don't really care about - the studios don't really care about Linux or the platform, they care about making good movies, they care about special effects and doing great work. So the infrastructure is just a means to an end, and whatever it takes to solve that as cheaply as possible, and effectively as possible, people are going to do.
DJ: If your primary business was selling software, for example, you might have a completely different attitude to it. That's not your primary business; you are users and developers of software rather than salesmen. What were the first applications to be shared, in terms of chunks of code, between the companies in the industry?
Doug: From Digital Domain's point of view, we have Bill Spitzak here, who wrote Nuke, and he's been here for about twelve years. He has constantly been working on this compositing program, and when he initially started doing this, he said he needed a fast user interface toolkit - and he put together FLTK. At the time he just asked 'Can I open source this?', the company said Yes, and off it went. So that was our biggest initial open source thing. Open source is difficult though - to put something out there, that you want to be out there, it takes more work. We have lots of math libraries that I've often thought 'Hey, this would be nice to share around the world', but all those math libraries are intimately tied with other parts, and so it's either that you put it all out, or you don't put anything out. If I were to take the time to separate it out and to make it a compact, sendable-outable thing, that's a lot of work.
Dave: A lot of legal hassle too, unfortunately.
Doug: Unfortunately, that has come up. Back when Bill did FLTK, the company was just forming and they were like 'Yeah, sure, whatever, we don't care.' But FLTK alone has been taxing for Bill. He's stepped back from supporting it, he's got his own copy now.
Dave: OpenEXR is the other one that comes to mind, and it seems to have been very effective so far too, in terms of solving problems and getting out there. People are familiar with it.
DJ: The new version of CinePaint is moving to FLTK, and they already use OpenEXR.
Doug: CinePaint is one of those out there; we keep looking at it.
Dave: It's not quite what we need yet, so we go back to where we are.
DJ: What would you use instead, for that sort of retouching task?
Doug: We use Photoshop.
DJ: Do you run it on Linux with WINE?
Dave: We don't, we just run it on Windows. Because we're one of the few places that dual-boots Windows and Linux everywhere, we tend to get away with a lot more, versus the complete integration that a lot of other studios have to face: 'How do we get everything to work on the Linux platform?' because there is no alternative. We don't tend to use a lot of Wine, because we've got our own Windows emulator that's already on there - called Windows! We've talked about it on and off, because other people have had great success doing that.
Doug: As a software developer, being cross-platform is both a blessing and a curse. Oftentimes you'll find a bug on one thing or another which makes your code stronger, but it's a pain in the ass - it would be so much nicer if I could just stick with Linux. When we write code, we have to make sure that everything runs on both platforms, so threading is always a little bit of an issue; there are differences in how things are threaded. We use Qt, it's nice, and it runs everywhere. Any libraries that we get, some math libraries that we get from the outside world, have to work on both platforms, otherwise we just can't use them.
Dave: Me too - more of a curse than a blessing. We tend to have to settle for a lowest common denominator because of it, versus saying 'Let's just do it right, and be done.' Especially with the Windows side, because with Windows when you have a problem, you're mostly stuck - you can't go fix it, you can't get the software guys to help find the bug in the code. For us, it's been a pretty big burden, but at this point we've solved a lot of the integration issues that cropped up with working on both sides. If someone had to start up from scratch and do what we did...
Doug: It took us ten years...
Dave: I couldn't imagine trying to get to that level of integration. But now that we're here, it's easy to continue onward, and get the benefits of having a really good 'Windows emulator' installed everywhere, and have the cross-platform technology working.
DJ: What's the administrator load across the platforms? Does Windows need as much, less or more attention, compared to the Linux stuff?
Dave: Windows probably needs more, really because Windows also tends to be the place where all the third-party applications are. We get a lot more random 'Here's a cool XYZ app that we want to try out, let's see what it is.' So we're installing the new third-party application and dealing with all that hassle, where as on the Linux side it's pretty much the core apps and internal tools, that we don't have to worry about.
Doug: And drivers - Drivers for graphics cards over the last ten years have been a pain in the butt.
Dave: I think that's improving. We used to be an ATI shop when I started; two years ago we switched over everything to nVidia with good success. We had a few rough patches in the transition, getting through the initial wave of problems and whatnot. Invariably, when we get a new kind of workstation which has the next generation of graphics card, there are another couple of bugs going on that we need to work out, because the drivers don't quite match up - but they've been very good for us.
Doug: Over the transition from SGI to Linux, the systems guys had their own hell to deal with, but from a software developer's point of view the graphics cards were just behaving so oddly sometimes. We were pushing them - we weren't writing games, we were writing apps that used different things than the games people used on the graphics cards.
Dave: They had a lot of catching up to do, because the Linux driver started well behind, especially on the enterprise or high-end graphics workstation level, where it wasn't a game thing, it wasn't running Quake or Doom or whatever. It was really pushing the pixel shaders. As the third-party vendors - Alias, Maya - had to work out their issues as well, the various graphics card vendors really got on board to fix the issues.
DJ: Although the games market may be much larger, because you are so high-profile, and you are pushing the technology that much further forward, they can sell you bigger and faster things every year...
Dave: We're the ones spending a thousand dollars on a graphics card, which your average home user is not going to be doing.
DJ: You're using nVidia's own drivers?
Dave: Absolutely, on both sides. They're working very well. Every time we have a new graphics card, we have that crop of problems, but we work through them fairly quickly with them. We tend not to have anything new pop up, because for the most part, all of our third-party software is tested against the nVidia drivers as well, usually before they get to us.
DJ: Would you say the nVidia cards are pretty much a de-facto standard in your industry?
Dave: For the most part - I think a small amount of people use ATI, but I'm not sure about the numbers from other companies. A good portion certainly use nVidia, and those are your two main choices.
DJ: Which GPU's are you using on your artist workstations now?
Dave: The older workstations use the Quadro 700's and 750's, up to the FX 1000's, 1100's and 1300's. The FX 1400's are the latest and greatest. A couple of the machines use the FX 3400's for some specialised things, because they're really high-end cards.
DJ: Have you done any work using the GPU for more general tasks?
Doug: Yes indeed, we've written a couple of things. Colour correction on the GPU is a big thing, being able to adjust the colours in real time - we've written a pixel shader for that. Just recently, we've taken a computer vision application called Optical Flow; this is a technique to figure out how all the pixels in one image move to the next image in a sequence. So you're watching the motion of pixels; not in 3D, but in 2D. If you can figure this out, there's all sorts of stuff you can do - compositing wise, computer vision wise. But it's a big problem - you have to figure out where every pixel goes in the next image, and there's zillions of pixels in an image. So that cried out for some GPU acceleration, and we had a project last year that accelerated that. It was a very interesting project; but the graphics card we were developing on was an older one, it wasn't PCI Express. Because you're pumping a lot of data, you're not just sending small textures to the GPU, you're pumping the entire 2K image twice, actually over and over again, the data path got a little boggy. But we did get about a 2x speed-up fairly easily, and we're looking at making it faster with the next bump in hardware.
DJ: Is it true to say that GPU programming is kind of obscure?
Doug: Duh, yeah! It's difficult - we don't look to the GPU initially, we program to the CPU. Once we get it working and we see an area where something like this is being used over and over again and could really use a boost in speed, we'll address it. So we're looking at fluid simulation on the GPU, but that's really difficult, and the Nuke compositor is also looking at the GPU to do some basic things that the GPU is aimed for. But you have to rethink exactly how to do it all.
DJ: What I'd heard was that it is an approach that is potentially very beneficial, but extremely difficult.
Doug: Yes! We had an intern, a software developer working on it for about three or four months, and it was their first big GPU project, so there was a learning curve going on. They were exploring and learning all the limitations by experimentation.
DJ: You mentioned using 2K pixel resolution, but I'd heard that 4K was being used more these days. Presumably with HDTV and digital cinema, there's an 'arms race' of resolution, and you're going to have to do things at a higher resolutions than you did before.
Doug: There's a big argument going on - in fact I think the Academy is holding meetings on what the appropriate resolution should be for the future. You don't want to go 10k - you don't need 10k, you can't see 10k.
Dave: The two questions really are: is it going to be 4k next, or should we skip that and go to 8k next? We of course are all hoping that we're going to stick with 2k for as long as possible, and then 4k for as long as possible, because it kills the rendering times, they go up four times; it's obviously squared, because you double one resolution and the amount of pixels goes that much higher. There have been talks, in terms of bidding projects, where they want to have 4k deliverables. Sometimes it's just a matter of up-resing, because it may just be we can do it 2k and then upres at the end. Sometimes they want full 4k stuff. It's really customer driven, when the customer says this is what they want out of the project, that's what we do.
DJ: Do you actually have a 4k display, or do you have to split that across four monitors?
Dave: There are some high-res 4k displays used for medical imaging. We don't have any, because we don't have that problem yet. We're starting to look at it though, because it may become a problem. They're quite expensive.
DJ: I suppose one of the factors is digital cinema, and whether the local cinemas start putting in digital equipment and throwing out the film.
Doug: It's a big argument, because you talk to the people who are actually generating the images, and like Dave said, a lot of times you'll render at a lower resolution because you know to match the film, you can get away with that - blow it up and blur it a little. Oftentimes, CG stuff is way too sharp anyway, so to make it fit you've got compositors that are blurring it. So if they're blurring it at 2K to make it fit, blurring at 4k is a waste - you're just blurring more pixels. If you don't do it, it stands out, and people complain 'How come I can see the CG? It's really sharp!' Bit depth is another thing that's going on - right now, we work in 10-bit log when we scan, and we do all our renders pretty much at floating point depth. How that's stored internally is 2k floating point images, and then rendering it back out, it gets compressed down to 10-bit log. So I know the same Academy group is looking at not only size, but depth, and how to represent colours.
DJ: That's why CinePaint got started, when it was FilmGimp, because of the colour depth problems of the software that was available at the time.
Doug: Absolutely - being able to paint in 16-bit is great.
Dave: One of the things that's driving higher resolutions isn't necessarily that there's digital cinema, and that we have projectors that can do this, but it's the long term process of 'OK, ten years from now we're going to want to re-release on the next-gen DVD or whatever, another version of our movie. We don't want it to come out all crappy because it's a lower resolution that doesn't make sense any more.' That's why the studios are all concerned that we want to have the right number, so it matches the display technologies of ten, fifteen or twenty years' time. Unfortunately that pushes the burden off on us!
DJ: I heard that one of the things holding digital cinema back is that although the distribution will be very cheap, the projectors themselves are expensive, and because traditionally cinemas paid for their own projectors, the studios can't buy projectors for the cinemas because of antitrust issues - as it would be like a subsidy to their own customers.
Doug: There is some sort of fund they've put together, some sort of organisation that's trying to put digital into cinemas, but maybe that's enough of a separation between the studios.
Dave: It's going to be customer driven - when people aren't going to go to movies any more because they don't want to see the old film projectors, they want to see the new digital screens, everyone's going to buy them, and that'll be the end of it.
Doug: It's also the cost of film, the physical film itself is very expensive, making those prints and distributing them. They're very heavy, there's a lot of chemicals and it's all icky. Once the distribution comes in when you can just distribute over a satellite to wherever, that will be a cost benefit - but that's long-term. Once we get there it'll be cheap, but to get there will be a huge climb.
DJ: It used to be that films would come around quite slowly - in England, we'd get them a year or six months after the US, and then they'd travel around the cinemas. Now there's a lot of pressure to release films pretty simultaneously. If you go to see a Harry Potter, it's out at the same time in the UK as it is in the US. We hear a lot of complaints about piracy, but from your view is that affecting the industry?
Doug: We're very separated from that legal stuff. The only thing I've noticed is that studios seem to be a little tighter on security while we're working on the film. To have images reach the internet beforehand - they seem a little more anal about that.
DJ: Because you see the film before the studio does, presumably...
Dave: We only see pieces. We sometimes don't see the full film until it's getting released. And we don't hear the audio. Sometimes, like in I Robot, we did a bunch of lip-syncing stuff, so we had to have the audio to have the robot talk properly.
Doug: But it was just the voices, no sound effects.
DJ: I suppose it's like in the old days of car manufacturing, when somebody made nuts and bolts, someone made the seats...
Dave: We did the paint job!
Doug: Rather than piracy, I think the main thing that affects Digital Domain is the maturity of the industry. People can set up an effects shop fairly easily - buy some Linux machines, run Maya on them, and you can start doing effects with ten people who get together for a weekend. So there's a lot of competition - there's the big boys, and you've had a couple of new people enter the market like Weta Digital. But there's all these small groups starting out, with ex-employees of ILM or Digital Domain, and those compete in some of the lower-end markets. Some stuff that we bid on, now we're too expensive, because other people do it cheaper and smaller.
Dave: The tough part too is that people try to buy the work to get them going, and then go out of business a year later. It doesn't help us at all, unfortunately.
DJ: The technology is extremely accessible, isn't it? Obviously, you've saved money by moving to generic PC's rather than SGI, but that now means that anyone with a PC can get started. At one time an SGI would have been out of reach for a small team of people, just the cost of the equipment alone. Now you can see documentaries that are practically put together on a camcorder, and people are buying them on DVD. That must be a concern, when you're making these really big-budget action movies.
Doug: The big-budget action movies are still a big draw, and they still sit on top of the box office. You need a big, concerted group of people to do the effects on these things. It used to be that a big effects movie would have a hundred or two hundred shots. Now big effects movies start at six hundred and just go up. You can't do that with ten people in a reasonable amount of time - you still need the bodies to just crank through the number of shots.
DJ: That creates an arms race of its own, in that the next movie has to be even more exciting and even more action packed than the last one.
Dave: And more creative in terms of the effects work; people expect blue screens nowadays almost as a default. But doing fluid simulation work or hair simulation - those kind of things are where people get excited. Doing giant monkeys running around that look real.
Doug: We're just here trying to make something look cool. If you go to a film that you've worked on, and you're sitting in the theatre, when an effect comes on and you watch the theatre and they gasp, or you can tell that they're engaged - that's awesome. They're not being taken out of the film because it looks like an effect.
DJ: Would you like to plug your next movie?
Doug: We are working on three new movies - one is Zoom's Academy, with Tim Allen. We're working on Super Ex-Girlfriend with Uma Thurman and Luke Wilson - some of the effects are pretty darn funny, so it should be fun. The other big film we're working on is Clint Eastwood's next film, called Flags of our Fathers, about the guys who rose the flag on Iwo Jima in that famous photograph, and their lives after that.
Digital Domain www.digitaldomain.com
SIGGRAPH 2006 www.siggraph.org
nVidia Quadro FX video cards www.nvidia.com/page/quadrofx.html