 from Boston, Massachusetts. It's theCUBE, covering HPE Big Data Conference 2016. Now, here are your hosts, Dave Vellante and Paul Gillan. Welcome back to Boston, everybody. This is theCUBE's Silicon Angles. theCUBE and our production of HPE's Big Data Conference. Hashtag, seize the data. Jeff Weik is here. He's the animation CTO at DreamWorks. Jeff, welcome to theCUBE. Thanks for coming on. Thank you very much. So, we were just talking off camera about the really fascinating journey that you've taken from hand-drawn automation to where we are now with computer generation. But take us through, like, first of all, what you do at DreamWorks. Sure, I'm responsible for the technology that's used for the creation of our animated films, both at our studio that's in Glendale, California. We have another studio that's in Bangalore and we have a sister studio that's in Shanghai. Okay, and then, so your first film, where the credits were rolling with your name on it, was The Prince of Egypt, right? That's correct. You were telling us that those were hand-drawn animations. Yeah. So, take us back to how long ago was that and what's transpired in the last couple of decades? Yeah, it's been a really interesting road. I've been there, this is my 20th year at DreamWorks. I started as a software developer on Prince of Egypt, coming from Aerospace, which was an interesting transformation. Although, you know, technology problems are technology problems, you know, all trying to solve the same thing. But there's been so many changes in, of course, in computer animation and what you can possibly do with animation from going from the traditional to what we call traditional, which was a combination of hand-drawn but yet digitized images and then being able to add 3D effects and things like that into the films. In fact, Spirit, which was one of my favorite movies, was a hand-drawn horses, but there are elements in there when we did camera moves where the horses were actually 3D and we tune-shaded them, made them look 2D. So there was a lot of integrated 3D that was sort of invisible in those films. How has that changed the role of the animator and the storyteller? I mean, a lot of this basic blocking and tackling they don't have to do anymore. Well, you know, the animators are really the actors, you know, so they are creating the emotion, you know, all that feeling that you have in a character and whether it's hand-drawn or it's more 3D animation, I mean, the tech, the talent that they have and the techniques that they use are really the same. It's how do I make that character remote? And we do have a lot of animators that started at DreamWorks as traditional 2D animators that are doing 3D. And we also have a lot of animators that draw over, they'll use digital pencil on a screen to be able to create the blocking and to be able to do a lot of that in a more natural way. We hear a lot about companies struggling with the need to train people to think digitally. They've been used to doing things the same way for so long. You experienced that at hyperspeed because your industry changes so quickly. How have you brought people along and trained them in those new skills and really got them on board? We actually, when we went and made the decision to go from traditional to full 3D animation, we trained our animators, you know, we had classes on campus to train them to use those new tools. But it's really a tool change. It's not a mental change. So how are you using analytics? Let's get into that a little bit. This is an analytics conference. Where does it fit? Yeah, it's, you know, I did a session this morning to talk about, you know, the big data analytics at DreamWorks. And if you think about animation, you know, the way that it's done today, it really is digital manufacturing. So every single, you know, thing that's created that you see on the screen, whether it's, you know, your eyebrows or dew, fog, the characters, the entire environment, everything is created on the computer. And there's probably half a billion files at the end of a film that get created that have dependencies in them. And it's that all those different files that, you know, includes models and textures and animation curves, effects elements, all have to be combined in rendering that final image. So if you think about all of that data and what the dependencies are, being able to track and know where all of that data is and be able to stage it to where the artists are, only allows them to not sit around and wait, but to be able to work immediately. And that drives efficiency into the animation process. Okay, so take us back. Like, when did, I mean, it's always been data. But when did the sort of analytics journey, you know, begin and how did it begin and how has it evolved? Well, there's a couple of ways that we use analytics. So I mean, one is in anomaly detection. So believe it or not, despite my best efforts, sometimes things go wrong. You know, when you're creating a film and you'll find something that's, you know, for whatever reason, you're either underutilizing your resources or an image isn't coming out right or it's taking too long. There's a lot of that, those problems that happen. And even as late as four or five years ago, after a film was over, we'd have these problems. And it's usually when you can least afford it. It's sort of at the last few months of making the film when everybody's working on it. You've got the entire studio focused on delivering that film. You'd find a problem. And what we would end up doing, and I remember a couple of years ago, we printed out all these log files on these big, giant sheets of paper and we covered the conference room with all these log files and we're going through with highlighters trying to find what went wrong. Again, this is after the movie's already done. There's nothing you could do about it. But we're trying to analyze it for the next film because every six months we have another film that's going to go through, you know, from crunch to crunch. So what we were able to do is actually when we first brought Vertica and we were able to not crunch things, you know, after the fact. Again, we, you know, but we started throwing things into Vertica and being able to do that analysis. And it allows you a couple of things from an anomaly detection standpoint. We had an issue on Kung Fu Panda 3. I'll admit it. We had issue where we had what we call low CPU utilization which meant on our render farm which right now we get about 85% 24-7 utilization of our render resources. And I'll talk about how we improved that by 15% with Vertica. But we had an issue where we were about 60% which is low and in fact it's so low that, you know, we didn't know if we could finish the film because we needed every bit of that rendering resource to get through. But we were able to go in and actually look in almost real time at what was exactly happening. What it turned out was our scheduling software was actually running longer than it was taking a whole bunch of short jobs to run. And so we were actually starving the farm by trying to schedule these jobs. And so what we were able to do is adjust our scheduling software to be able to accommodate these, do that immediate adjustment. And within about four hours from problem to solution we're able to get back up to the 85% utilization. And these problems are problems that would be visible to an untrained eye or this is like the wedding where something goes wrong and nobody notices. Are you guys just like ultra perfectionists seeing problems or is this like a really visible collar issue or a blip in the screen? No, no, it's actually, I'm not getting my work done. I mean, what's happening is, you know, we have these huge batch- You can't finish. All right, you can't finish. You can't, there's a window and you're not going to hit it. Right, the next, you know, we submit jobs the night before the next day when artists comes into work, they have reviews and they, and you say, yeah, your image isn't here yet. And now what do you do? And that's pretty serious stuff. Now, how has technology evolved, whether it's analytics related or not, just generally over the course of sequels? Like, you know, Shrek one to three, how much did the technology change? Did you have to start over, you know, from scratch for three or how does that all work? It's crazy. I mean, you think that with a sequel, you'd say, wow, we're going to gain all this advantage, you know, because, you know, we already have the characters built, they're already rigged, we all have textures, but technology changes so drastically and there's so many efficiencies that you can gain now that we actually, it's kind of a do-over. Now, one of the things that we don't do-overs, we know what the characters look like, we know how they move, we know the emotions, we understand them, and that's, that part of the creative process is very time consuming to really get that feel and that personality of the character. So that and some of the locations, you know, we already have those identified, but we will go in and actually, you know, we will re-rig those characters and re-texture them and take all those advantages that we can then, you know, get in lower rendering and more creativity. I'm sure the audience of standards get higher as well. They do. Well, our standards are probably a lot higher than the audience of standards as well. Can you talk us through a minute of video? What goes into a minute of animated video? Well, there's 24 frames or images in a second of video, so you can do the math. 1440, yep. Hey, good job. So there's that, but what is in those images? There's, if we talk about the backgrounds, there's a location that you're in. So the desk, the computers, the, you know, all that sort of thing that has to go in. What's in the background? There's us, you know, so you have to create us as characters that have to be modeled. Our character effects, our clothing, hair, glasses, all that sort of stuff has to be created. The animation that we're moving, the lip sync to go with the actor's voice, that we have to match, that has to be done. And then at the end of it, we put all that together and, well, effects, we won't have any today, I hope. But then you take the lighting. So all of these lights, which are digitally generated, that then have to, when we assemble all that, bounce off of our skin or go into our skin and subsurface scatter with the geometric procedurals that take the guide hairs and fill in all the other hairs. So all that stuff has to happen, the clothing simulation as we move and stretch, all that has to go into the spinal image that thing gets rendered. How about, it seems like as time goes on, there's a tighter relationship between the animator and not just the voice, but the personality of the person behind the voice. It seems to come to life. Is that just my imagination or is there deliberate attempt to, first of all, find voices with actors behind them that reflect the character that you're trying to reflect? Or is that just my imagination? No, it's actually deliberate. When we pick a voice actor to, and by the way, that whole casting process, you want to get an actor and a voice that really fits the personality that you want that character to have. Then you have to really, you have that voice. And by the way, the voice is done first. First we use temp voices, but then we get the actor's voice. They record that first and the animators will animate to that voice. But they also, we videotape the actors when they're performing so that you actually get that as a reference to start from. Now you exaggerate it, of course, and that's part of the animator's acting job to make that exaggeration. But even like, Toothless and How to Train Your Dragon was really modeled after Simon Otto's cat. He studied his cat and how the cat walked, how it interacted with him. So you use a lot of those references and animators, they actually have acting rooms where they go into and they videotape themselves, whether it's a sword fight or it's some emotional experience, they'll do that and use that for reference. And then they have mirrors on their monitors so they can see what that looks like when I open my mouth or I crunch my eyes. So what is something else? Kung Fu Panda was you, right? Yeah. Was that right? Madagascar. Madagascar. Be careful, be careful. I know, it's always true. Okay, yeah. I've been careful. About 32 films we've done so far. Saving Private Ryan was not animated. It was not animated. That would have been a really tough kid's film. Was ants you? Ants was us. Okay. That was our first film actually. Three for three. There you go. I don't know what my twin five-year-olds are doing with that. I'm going to stop there. All right, great stuff. Anything, sort of here that you've seen or you mostly conveying to your peers, I mean, stuff that's interesting, things you've learned, takeaways that you're trying to get from an event like this? You know, it's always great to hear how other people are using data to be able to advance their own business. And we're always open to hear what's coming and what's new. One of the advantages we've had with our relationship with Hewlett Packard Enterprise is being able to get advanced roadmaps on what's coming and to be able to work with labs to find out what technology is coming. We're working with them on the machine right now to look on how that can potentially change the way that we make films. And that innovation is such a heart of our company and Hewlett Packard. And I think being able to innovate together, it really gives us a competitive advantage. So I'm excited to learn what other people are doing. Actually, you guys have a long-standing relationship, the other Jeff, too. We do. So thanks so much for coming on theCUBE, Jeff. Really appreciate it. Great, thanks very much for having me. You're welcome. All right, keep right there, everybody. Paul and I will be back with our next guest right after this short break. We're live from Boston. This is theCUBE. We are here to win championship.