Christopher: Hi Curvin, thanks for making time to talk today. What I thought we’d start with is your background. How’d you come to visual effects, and teaching Houdini?
Curvin: So I sort of straddle the fence between technology and art, for sure. Since I have a bachelor’s degree in sculpture, which I think is what really lead me into 3D, and then I got a masters of science in interactive technology, and then I started working in the field… actually, I come from training and simulation more so than arts and entertainment. I did simulation work for the government mostly when I first got out of grad school. So we worked on some simulations for the Department of Defense, National Security Agency. I worked with the Center For Disease Control. A lot of stuff like that. And they’re like video games, except they’re real world situations. And then I also went back and got my MFA in Illustration. So I have a masters of science, and a masters of fine arts. So that’s sort of my training background. But now I’ve been teaching full time for about 20 years. So what that has afforded me is the ability to kind of — I’m sort of a jack of all trades. I’ve learned just about everything. I started in modeling, then I moved to animation; I’ve run the gamut really, which is great. And I’ve worked with nearly every 3D application, Max, Maya, Rhino, AutoCAD, Cinema 4D, you name it.
Christopher: Right. And was there something specific that took you where you are now?
Curvin: So where I am now, Becker College — I’ve been [a professor] at Becker for eight years. Prior to that, I was at DeSales University in Pennsylvania for four years. And I actually started teaching at a community college in Pennsylvania, that’s where I got started; so I was a department chair there for eight years also. I came up to Becker about eight years ago, and that’s where I really hopped on the freight train for all of this stuff.
Really, it’s just been in the past two years, I would say, that I really started to come towards visual effects. And that started with Houdini. About two years ago, I went to Boston FIG, and I had this really great student who was there also, and he ended up talking — there were a couple of Side FX people there. He talked to them, and he was like, Curvin, I’ve got to introduce you to this guy. So I went over and I met him, and boy, I forget who it was now, that’s really unfortunate. But all of a sudden, we’re like, oh my god, we should do Houdini. So I went to our department chair and said, hey, we should probably jump on the Houdini bandwagon, and he said okay. So I was the guy who sort of lead the charge, and that’s where it took off from.
Christopher: I think you talked about — or we touched on it, Houdini, it’s kind of — you’ve got to put some energy into it. It’s not like Octane, or Redshift, where you have very simple UI…
Curvin: I know. Octane is like, one, two, three, click; look how great it is.
Christopher: Exactly, right. It’s doing all the heavy lifting in the background. [Houdini} is kind of like a manual transmission. It’s like a manual everything; manual transmission, no power steering. But, you get these amazing effects, and maybe that’s one thing — probably I think the place to start with it is, how — in what way Houdini stands out in the pantheon of 3D applications. What has made it distinct, and then, maybe we can talk about strengths and weaknesses. But what has Houdini be distinct among Maya, Max, and the other apps?
Curvin: I think — we’re going to talk about its strengths, which also end up becoming its weaknesses, in terms of learning. But yeah, its strengths are obviously — you hear it all over the place is that it’s non destructive, and procedural. You just can not beat that. And the fact that you can build a pyro simulation, and then go back and decide that you’re going to use different models at the front end of that simulation, and just plug those nodes in, is just phenomenal. That’s amazing. When I come back to Maya, and work on some effects, I just feel like my hands are tied. It’s like, okay, do I want to make this commitment, okay, press okay. Because you know that if you want to change something, you nearly have to start over in many cases.
Christopher: I guess maybe that would be a good thing to distinguish – how does this compare with 3dsMax?
Curvin: I think Max actually has a little bit — I feel a little bit of — I feel it has a little bit of non-destructiveness to it with its modifier stack. I used to teach Max quite a bit, and I also pointed that out, like, look, the modifier stack makes Max more or less nondestructive. But I think Max has really had a good stronghold in mechanical and architectural visualization, and those areas. I’ve worked with a lot of those kinds of companies, and I always go to Max for that. Because you can just — the data just comes in from Inventor, and Revit, and AutoCAD so well from Max. But I think on the visual effects side, Max, as an off the shelf solution, has always been playing catch-up. It’s had fantastic plugins you can buy to make it a Maya, or maybe even a Houdini. But off the shelf, it’s always been playing catch-up. So I think that is one of its weaknesses is that if you want to build a really amazing explosion, let’s say; it just didn’t have the volumetric pyro tools to do that for the longest time.
Christopher: Coming back to Houdini’s strengths, are there other specific advantages that you — that people highlight? I just love this about Houdini, that I can do X. Is there anything else that comes to mind?
Curvin: I think just the open-endedness of Houdini. When you open up Houdini, it’s like a blank canvas, and you can construct your solution any way you want. You can build it all inside one geometry node, or you can spread it out. It’s wide open. So it’s like, you want to drive somewhere in a car. Well you’ve just got all of the engine parts and you can decide how to put the car together.
Christopher: What about weaknesses in Houdini, that — maybe even comparing using it — comparing it with another 3D package?
Curvin: I think really Houdini’s one and only weakness is its learning curve; I mean it’s a beast, and it’s an absolute beast, and it’s a beast because of its strengths. It’s so open ended that there are an infinite amount of ways to come to the same solution. And the challenge to learning Houdini with that in mind is that — you can go online and find all kinds of training videos on how to build a pyro simulation, let’s say. But you will also find a hundred different ways to build the same pyro simulation, because it’s so open ended. So I think one of the challenges is you have to realize that when you watch somebody go the process of building, let’s say a pyro simulation, in Houdini; that is not the only way to do it. There are a lot of other ways to do it too.
So it’s the learning curve. And I think the pitfall that a lot of people fall into, and I’ve seen this with — and I’ve heard it from other people anecdotally, and I’ve also seen it myself — is that when even an experienced effects artist doesn’t know Houdini, when they see you do something in Houdini, it looks amazing. So the first knee-jerk reaction is that, well it must be easy to do, because Houdini can do it so well. But what they don’t realize is all of the blood, sweat and tears you put in to learning how to do it, to get there. So they sit down in front of Houdini and become frustrated very quickly.
Christopher: Yeah, yeah. But it makes something, sounds a little closer to a classical art, where an oil painter is not going to achieve the effect that a master oil painter has with the years of training. Understanding what you do with the oils, and how to mix them, and the brushes that you use. Which affect the kind of light on object experience that has, and had, oil painting be such a magical experience, in previous times. Right? It sounds…
Curvin: Right, right. I think I would add on top of that, using that analogy, that Houdini would be like, say, okay, well here are all of your minerals, or whatever you want to call them, that make the colors. You need to figure out how you want to combine those before you start painting.
Christopher: Well, I think that brings us to another good topic, and that is — maybe you, I think, would be a good person to ask. What Houdini users struggle with? Like what they might say is, if Houdini only had X, is something missing in Houdini that people regularly complain about, or wish it had, that you see and hear about?
Curvin: I can’t think of any certain pattern I’ve seen out there on like, where there’s one major thing that Houdini needs that it doesn’t have. I mean, I can say from my own experience, because of Houdini being so open ended and non destructive, in some cases, it just makes simple things really difficult to do. Like you can’t really just bring a file in and convert it to an FBX. You can, but there are so many steps involved that it’s easier to do it in Maya. But I don’t think that there’s a solution to that. I mean if there was, you would be compromising the foundation of Houdini.
I’ve done a lot of character animation in Maya and motion builder. And I just haven’t tackled it in Houdini yet. So I don’t think I could say, Houdini, that’s where Houdini is lagging. And I also wouldn’t say that that is a major point that I’ve seen in forums out there either. So yeah, other than just learning Houdini, I have not seen any real technical pieces where anybody has said, this is one thing that it’s lacking. I think that the developer that Houdini — or Side FX, when they build a new tool, they just do it really well.
Christopher: It does seem like there’s a rigorous process underneath. And speaking of processes, hardware specifics. We’ve done some benchmarks, and it’s kind of a CPU versus GPU world in Visual Effects and CG. From the work that we’ve done, it’s clear that Mantra, the built-in renderer, is CPU based. And there are functions within Houdini which are more GPU reliant? Can you speak to those differences? There’s Pyro, there’s —
Curvin: Oh, there’s like RBD, Rigid Body Fracturing, Flipfluids, all of that stuff.
Christopher: Yeah, Flip is another common one. We ran benchmarks with some static — some basic models, Flip, Pyro and others. In some cases, Houdini is using all of the CPU cores you can give it. And in other ones, it’s not. It’s not linear or absolute. A lot of cores is generally good, but frequency is important too; so you see both come into play. Are there functions within Houdini that you’re aware about, which are really just purely CPU, others which are just GPU reliant?
Curvin: Well, that’s a good question, and from my experience, I’m really nebulus on that. I don’t know. I’ve been trying to figure it out myself. Other than official statements like, Pyro is GPU based. Outside of that, I can’t say from experience how that is. What I find is that the general workflow in Houdini is to cache stuff out, and cache it, and cache it, and cache it; which means that you need a huge amount of storage space. Like for instance, I did a prototype piece for Billy a few weeks ago, and we set the passenger seat of a car on fire. And because the Pyro simulation was so big and so fast, I had to increase its voxel field, so that it didn’t clip. And in doing that, my Pyro caches were approaching like, 5, 6, 7 gigs. So I know that’s one of the foundation principles of Houdini, that you always want to cache your simulations out, then you move on to the next layer. Now, whether they’re being cached GPU, CPU, or a combination of both, that’s debatable, at least to me, because I really don’t know — I haven’t sat down and done benchmarks for one or the other. The system that I run off of is I have a 16 core threadripper, and dual GTX 1080s, and I think 32GB of RAM. It’s still — some simulations can take 10, 15 minutes, to cache out.
Going to the Mantra side of it, that one, yeah, Mantra is CPU based. So with every 3D application, I do really like whatever their native render engine is, simply because it’s obviously the most integrated with all of the tools and functions of the application. And that’s what’s great about Mantra. But, Octane is insanely fast. And also… what’s that?
Christopher: And photoreal.
Curvin: Yes. And I think they just announced the Unreal Engine plugin.
Curvin: Yes. I’ve spent about the past month or so working with the (OctaneRender) developers on bugs – well, this doesn’t really work well, can you do this instead. I build my simulations, and typically I’ll go to Octane instead of Mantra for GPU rendering. But I think Mantra is a capable engine, but they all are. Arnold is great, Redshift is great, V-Ray is also an awesome one.
The one that I love, but it irritates the hell out of me, is Renderman; simply because Renderman is always just such a bear to try to get up and running in every application. I’ve got it up and running, and I’ve done it in Houdini, but the workarounds you have to do to get it to work is just crazy. So yeah. But I think they all have their — I think they’re all fantastic. It’s just — you can debate which one is quicker. Octane is known for being super fast. But if you find a really good Arnold render specialist, they’ll tell you that they can optimize Arnold work just as fast as anything else.
Christopher: To your point about caches, before we finish up on render engines. A fast, dedicated cache drive can have a real impact on workflow speed in Houdini I was speaking with an M&E specialist friend at the Kingston booth at SIGGRAPH just 90 minutes ago and he said the same thing.
Curvin: Yeah, yeah. And actually, I’m wrestling with that now myself, because my computer — I have an M.2, it’s only 500 gigs, and that’s my primary drive. So I’ve got it half filled with software and the operating system, and I want to run my caches off of it when I’m working in Houdini. But because the caches become so large, I’m like, oh well, I take the hit, and I run them off of my SATA solid state instead.
Christopher: We should run some tests, just to see what kind of impact a very fast dedicated NVMe cache drive has. And obviously you would need an open PCIE express lane, an X4, on the motherboard. But I would be very interested to see what kind of impact having twice the disk speed for your cache folder might have. Something totally dedicated, no applications, it’s just project assets.
Curvin: Right, right.
Christopher: So anyway, something we can explore.
Curvin: Yeah, definitely. I’d be interested in that.
Christopher: Coming back to render engines for Houdini, it sounds like you’ve got some experience in that regard for sure. Perhaps primarily Octane rendering for Houdini, but Redshift and Arnold DCCs are using as well, though I hadn’t heard anything about people using Renderman. Do you have any other comments about how Houdini plays nice — or not — with other render engines? What render engine is fastest with Houdini? Which render engine works really well, and which don’t?
Curvin: Well, not so much. So I work with nearly every render engine, but within Houdini, I’ve only used Mantra, Octane, and very briefly, I did some work with Renderman. So for instance, Redshift, I actually worked with Redshift and Softimage. Which actually, Softimage is one of the reasons I got Houdini, because I loved Softimage. That was the closest to nondestructive that we had, prior to Houdini. And I did a few projects in Softimage, and Redshift. Arnold was obviously Maya based, and then in 3DS Max is where I did most of my V-Ray work. So I can’t really speak to them and Houdini. I can say that both Renderman and Octane are a little clunky to get installed in Houdini. I’m not sure if let’s say Redshift is the same way or not, but if you want to get them up and running, you’ve got to get into Houdini’s environment file, and set up the proper paths, to where all of the files are located.
Christopher: Well, thank you for your time today Curvin. We’re planning to break our coverage of Houdini into two parts: First the basics of Houdini, including Mantra, then the different render engines, and include benchmarks with each.
To close, I also want to ask you what’s emerging in the world of visual effects, especially as it applies to Houdini? And how [do] those technological developments support your goals and aims?
Curvin: Yeah. So what I see is that the results we pull out of these 3D applications are becoming closer and closer to real time as time goes on. And I think seeing what EpicGames is doing with Unreal Engine in particular is just showing us that that path is opening up. And I know Unreal Engine has made a lot of inroads to the film industry now. They’ve moved out of games, which is fantastic. And then seeing an integration like Octane, into Unreal Engine, just shows that that line between the two is really starting to blur. And then on top of it, I can see at SideFX, they have been pushing hard with their integration into Unreal Engine as well. The Houdini Engine plugin, and then some of the new stuff that they’ve built, some of the new technology, to integrate with Niagara, which is the new particle and effects system that’s in Unreal Engine, it’s huge. It’s so big that I wouldn’t even want to really comment on it right now because I’m just scratching the surface; I’m learning what all is there. So that convergence of game technology, into film, and the render engines, and how they’re becoming more real time, is really an exciting thing. And I think that’s where it’s all going. Specifically, I really couldn’t say what we might see in the future. Of course everybody loves to talk about AR and VR, but I think those are two, I wouldn’t say struggling technologies — but people still want to sit down and watch a movie on their big HD or ultra HD TV, or they want to go to a movie theater. Putting on a headset, I don’t know when that’s going to happen, where that will become the dominant technology down the road.
I think as a researcher, I could push into AR and VR, but what I really see is, how do we take things like game technologies, and procedural technologies, like Houdini, and the render engines, and just push them out into other fields? We’re not just in games, we’re not just in — I work for an opera company in Boston, and we are using these technologies and pushing them to the brink, to put operas on stage. Which is a whole other avenue in itself. But I mean maybe it’s just me and it’s just my focus. But I like moving the technologies into existing art and design fields; as opposed to just looking at AR and VR as, how can we make this the predominant delivery medium for everything.
Christopher: Yes, not to mention practical applications like medical. Like a doctor being able to see a patient on whom he or she is going to operate in advance of the operation. Some of those applications just seem really valuable.
Curvin, this has been a pleasure. I will follow up with you re Enterprise NVME speeds and impact on workflow speed in Houdini next week.
Curvin: Okay, that would be great.
Christopher: Awesome. You mentioned, you’re going to Double Negative, and I forget the other place, in London?
Curvin: Yeah. So I’m going to Double Negative and Framestore, and then I’m doing a motion capture seminar at King’s College in London.
Curvin: Yeah, it’s going to be a lot of fun.
Christopher: Perfect, thanks again and you have a great night.
Curvin: You too, thanks Chris.