← Reshaping Workflows with Dell Pro Precision and NVIDIA RTX PRO GPUs
Reshaping Workflows with Dell Pro Precision and NVIDIA RTX PRO GPUs
Redefining VFX with JangaFX
00:30:40
Discover how JangaFX, Dell, and NVIDIA power real-time VFX workflows for film, games, and live events—explosions, fire, and more come to life instantly.
Loading summary
Transcript75 lines
- [00:05]
A
Welcome to Reshaping Workflows with dell Pro Max PCs and Nvidia, where innovation meets real world impact in high performance computing.
- [00:20]
B
Welcome back to another episode of Reshaping Workflows with Dell Pro Max and Nvidia RTX GPUs. I'm Logan Lawler, your host. Today we have something, I wouldn't say different, but a ISV that Cindy is very fond of. And you can see Cindy's back. I can't escape her, but are doing some really interesting stuff in the VFX field and use a massive amount of GPU compute, but we'll get all into that. So with us from Jenga fx, we have Jason Key. Jason, thanks for joining us.
- [00:51]
C
Thank you. Thanks for having me.
- [00:52]
B
Of course, ma'. Am. No problem. So why don't you.
- [00:55]
A
Where's my introduction?
- [00:57]
B
Cindy, please introduce yourself. You know what? Do it. Please tell everyone what amazing work you do.
- [01:03]
A
Cindy Olivo, marketing manager for Media Entertainment, and I'm the one who's partnered with Jenga Effects, not Logan. Just thought I'd throw that.
- [01:10]
B
That is true. I mean, I'll give it to you. But if you watch most of these episodes, even the ones Cindy has partnered with, also partnered with a la Logan as well. But anyways, Jason, take a second, introduce yourself a little bit on your background and give us a quick overview of JengaFX and the two products that we'll be talking about today. Sure.
- [01:31]
C
I'm Jason Key. I am a principal VFX artist and evangelist at jengafx. Just a little bit about my background is I've been doing 3D graphics stuff, I guess you would say, for about 30 years now. My educational background, my start was in engineering, so I spent some time in the CAD world for a few years and then eventually migrated over to the media and entertainment and worked in games and film for the past 20 years. So I'm bringing some of that experience and expertise to the Jenga effects team because all of their tools are involved with those industries.
- [02:11]
A
I would also say you're kind of a Dell evangelist too, Jason.
- [02:14]
C
Absolutely.
- [02:15]
A
Noticing the monitor in the background, we love that.
- [02:18]
C
Yeah, I really appreciate Cindy. She's been extremely helpful and generous in getting us access to all the latest Dell hardware. And so, yeah, we've enjoyed showing it off at trade shows. And I have a Beast workstation here in the background and another one on the way, so, yeah, it's great.
- [02:37]
B
Okay, well, tell us about the two, you know, main platforms or ISV softwares, Embergen and Liquigen. Give us a quick, you know, overview on what they do, how they're used, and what you do with them in vfx.
- [02:50]
C
Embergen is our tool for creating things like fire and explosions, steam and magics, basically anything that's a gaseous type phenomena. And then Liquigen is very similar. It's also a simulation tool focused specifically on liquids.
- [03:09]
B
Okay, so for someone like me that doesn't know anything about vfx, like, is this used to create scenes? Are you doing this to, you know, augment scenes? Give us a little bit more detail about those two platforms, kind of. I mean, not the special sauce like the IP of how they work, but how you would use them in a traditional, like media and entertainment workflow.
- [03:31]
C
Yeah, sure. So they're both standalone tools, so they're very, very focused. And that's one of the things that allows us to make them extremely fast and easy to use because they are standalone. But they're basically used to create VFX elements that you would export to other DCC platforms. So you would create an explosion or a splash of water and you would. There's various ways you can export that either in a 2D, you know, image compositing form. And we also Export industry standard 3D formats like VDBs and Alembic caches and things like that. So you create in our tools and then you export and integrate into the other tools like Unreal, Blender, Houdini, anything really.
- [04:25]
A
So, Jason, early on, you know, we had this discussion about, you know, when we were first kind of starting the partnership and you said something that, you know, was. Was really interesting to me when I asked you guys, you know, if there was any AI on the roadmap and you were like, not yet. You know, we are doing things with just pure brute force, which I thought was really funny. Is that still the case? Is there, is there AI kind of on the pipeline? You know, know, where are you at on your AI journey? As most softwares we're noticing, especially in this space are, are starting to integrate it into the software.
- [04:59]
C
So we have done some R and D in the past on exploring how some machine learning tech techniques could be used for fluid simulation. We're not currently using anything in our tools right now. They're still based on, you know, pure physics and mathematics. I have been exploring use cases, not necessarily, you know, built within our software, but how it could integrate on importing things or exporting things that have to do with AI. For instance, I was playing around with Meshi AI, which is a text to 3D program and generating some basic models of like a palm tree and import that into Embergen and set it on fire, for example. So things like that are things that I explore. But internally we've done some investigations, but we're not doing anything active right now.
- [05:54]
A
Oh well, if you need anyone to tinker, we got a guy, his name is Logan. Doesn't really have a whole lot going on and loves to be tinkering and solving.
- [06:03]
B
You know, it's funny that you say I don't have a lot going on because, you know, some Cindy recently started to have to call me team lead, which she absolutely hates. There's not team lead next to her title anyways, that's a whole joke. So it's really interesting that it's true. Kind of, you know, brute force, mathematics, ray tracing. I mean the whole thing, right, maxing out, you know, Nvidia RTX GPUs. So let's ask this like, you know, obviously depends on what you're trying to create, but let's say something traditional out of, you know, Imbergen. You're trying to create a fireball of a bomb exploding. Let's just pretend you know what type of. Obviously you have the highest end, you know, Dell Promax system. But for someone who's using it out in the industry, like what are kind of the minimum specs that you have to be at to run this workflow successfully?
- [06:49]
C
Our men's spec is probably would go all the way back to a RTX 1080 maybe. I think that's what I started on five years ago and I was able to do Embrygen simulations now they were smaller and lower resolution but yeah, you can go all the way back. So I mean obviously we recommend at least you know, the RTX generation, but you can go back even further than that. And Cindy mentioned brute force. That that kind of term kind of makes it sound slow. But the great thing about our stuff compared to other solutions in the industry is we're from the ground up built to work on the gpu like all on the gpu. So everything is built to be optimized and run on the gpu. So we're not even really using the cpu. And the CPU bound solutions are the ones that tend to be slower because they involve a entire caching and baking process that we don't, that we don't do. We bypass that when we're running everything in real time, which no one else out there is, is doing that right now.
- [07:58]
A
So Jason, for, for people listening that aren't Familiar with this space. You know, I see a lot of your demos on LinkedIn. You're kind of famous on LinkedIn and we aren't even talking about like pixels. We were talking about voxels. What's a voxel?
- [08:14]
C
So voxels like to think of them as 3D pixels, I guess. So instead of, you know, a resolution and width and height, you have resolution and width, height and depth, say. So if you really zoom in close to, you know, a simulation, you'll, you'll see little cubes, like little Minecraft cubes. Now you never, you'll never see those in your output because we're dealing with millions and even billions sometimes of voxels. So you don't actually see those. But that's, that's kind of the gist.
- [08:51]
B
A quick question like in terms of like let's, let's talk about, imagine for a second like how customizable is, you know, kind of the output. Let me give you an example, right? If you're trying to create a campfire with, you know, oakwood, it's going to kind of burn red, a little bit yellow, right? But if you're saying, hey, I want to create, recreate, you know, someone using a cooking stove for camping on like really high proof, like camping fuel that's almost clear, almost bluish in nature. Like can you control with an IM kind of like the turbulence meaning like, you know, if it's windy or not or like which fuel you're using. Does it get down to that level of specificity?
- [09:30]
C
Yes. We're not as physically based as, you know, you can't put in like propane and it will automatically turn blue. But we do have full control over the shading of flames and smoke. And so you know, you could control your, if you're talking about a camp stove or something, you could go in and make small flames and you could use for instance, say a gradient that you could create yourself or import to color the flames that go from like blue to red and the fall off. And yeah, you have full control over the, over the look.
- [10:04]
A
I know you guys started off as traditional kind of a VFX software. I know that you guys are, are slowly kind of moving into game development. What other industries are leveraging? Embergen and Liquigen.
- [10:18]
C
So our roots are games. So the founder Nick, he built Embergen was our first tool and it was built out of the frustration of existing tools being extremely slow and designed for high end film and for use in games. It was just overkill, overcomplicated. Extremely slow and Basically, it's like, there's got to be a better way type idea. So, yeah, we were built for games originally, but as we've optimized and improved and the quality starting to increase exponentially, combined with all this, the hardware advancements, we have a lot of people in film starting to use it and then, you know, all the other know, media, entertainment type stuff like advertising, live DJ shows. We've been used on the inside of the sphere, Grateful Dead shows. Yeah, it's really, it's really fun to see it grow because, because we are a very focused tool and much more approachable and fast. We're kind of, you know, people in the industry are picking it up that have, that are experienced, but we've also opened up this new market of people that would never be able to do it because the time investment and the expertise required just doesn't make sense for a lot of those applications. So we've got all that, which is great.
- [11:52]
A
So we've got you in our booth at Siggraph coming up and you're going to be demoing on the new Nvidia Blackwall card, which is 96 gigs of VRAM. What do you think the big differences are going to be versus what you're using now? I think you are currently on the 6000 ADA card. So this new generation is essentially doubling the VRAM. What is that going to do for the software?
- [12:13]
C
Oh, it's amazing. Yeah. Like, like you said, I've got the workstation I have here at home is. Has dual 6000 ADAs. But I did have the opportunity to work with you guys at both GDC and nab. So I have gotten my hands on the Blackwell a couple of times and it's just incredible. Like it's about twice as fast as the ADA generation as far as raw power, compute power that we need and then the double. The VRAM is super critical for us as well because I don't want to get too into the technical weeds here, but we, we've made the decision to stays a single GPU app, so we can't or we've chosen not to do a distributed solution that would spread across GPUs for technical reasons. It would. We'd have to sacrifice performance on single GPUs to make that happen. And it's also really hard. So my point is having a beast of a single GPU is extremely exciting for us. So the amount of VRAM that you have on a single GPU is essentially controls how much detail you can get out of our tools. So when you're talking about, I mean 48 gigs is a lot. When you talk about doubling that, you really get into like you know, feature film hero quality assets that we can, you know, know, simulate with billions of voxels.
- [13:48]
B
Quick question, like how? Like for example, I'm. It's a two part question. So before Iman or any competitor, you if you wanted a fire in the movie, I'm assuming you went out and created a fire somewhere and recorded said fire or blew up said bomb or we're at the beach for the waves. And now my question is part one of question is how much more are you seeing those liquid fire explosions type stuff not be done like in real life and then simulated through Embergen or Liquigen. And then the next part of that question is, you know, let's say you have a scene that is time lapse. So you've got people sitting around a bonfire, let's say for eight hours. You start fire is very small, it builds very big and then over the course of time, you know, it kind of dies down. Does Embergen actually take that into account like the time lapsing of the fire or does it really look at just that said moment in time that you're trying to capture?
- [14:52]
C
Those are both interesting questions. The first one, I guess I would. That's not really specific to us in general cg, you know, effects are mainly done where they make the most sense. So a lot of times you try to capture things in camera if you can, like if you can do it on set and most people like to do it that way. There's some places where it doesn't make sense. And so when you get into, you know, post production and there's a number of things you can do. So you, you can use stock footage like something like action vfx. You can get, if it's a super generic thing, you can actually use stock footage footage and, and scatter those in different places and then you get into the cg, you know, simulated stuff and that's usually where you want to really art direct something very specifically so. Or if it has to have interaction with a character, you know, coming through a fireball that you're obviously not going to be able to do that on set and you're not going to be able to get that out of stock footage. So that's when you're looking at custom simulations and yeah, that's kind of the basics of that. And then what was the second question? Yeah, you could easily do that. You would probably, you know, you would animate the fuel over time to make something die down like that. But yeah, we have full control over the combustion attributes. That's the one thing I really like about our tools are that they, they're focused and they're simple to get started, but they are not toys. They like, for instance, Enbergen I think has nearly a thousand parameters, so it's easy to get started, but you can also go very deep. So for, for instance, what you're talking about there, we have controls for fuel consumption rate and smoke gain based on fuel consumption and all kinds of little controls that you can go in and tweak if you want to go deep. But we also, you know, have really great defaults and we have a lot of free presets that come with it. If you're not wanting to go deep, you can just grab an explosion that we already have and start with that and tweak a couple of numbers and you're done.
- [17:10]
A
So Jason, so how is like, you know, for example, I was Enbergen, how does Embergen compare to traditional sort of tools like Houdini in terms of accuracy and control? Like, you know, is it another software? I mean, I know that it all kind of works together, but I mean some people have been doing this type of work and unreal or in Houdini or something. So what is, how does, how is Embergen different?
- [17:34]
C
The big difference is, like I've said, it's very focused. So you know, Enbergen does fire, smoke and steam, things like that. And Liquigen is focused on liquids. Some of the other tools out there like Houdini are, are amazing and, but they're a much, much broader scope and what they can literally do anything you can im. So in our tools, you're not going to be doing the modeling and the texturing and the character animation or crowds. It's very focused to do one thing really well, really fast. So that's the biggest difference. And again, being built from the ground up to run, optimized exclusively on the GPU is a big differentiator because when you're able to see the results in real time, it allows people to iterate so much faster. Like so traditional simulation tools, you for instance, like you said, Logan, if you want to change the fuel rate, say you go and change that number, you don't see what the results are. You change that number and then you wait, you wait for it to cache, you reload that cache and then you have to render to see like what it really is going to look like. So we're not only simulating in real time, but we're also rendering in real time. So you're getting, you're seeing like the blue flames fading to red. All that stuff is being done in milliseconds versus minutes or hours.
- [19:12]
A
So I've seen you smacks out those GPUs like on, on the ADA. And we're excited to see what can happen on the Blackwell. But what are the limits? Do we know yet?
- [19:24]
C
The limits. I'm not sure what you mean by that.
- [19:27]
A
Well, I mean what are the limits to the complexities that Embergen and Liquigen can handle?
- [19:31]
B
Yeah. On a 96 gig Blackwell with 600 watts, what is the biggest thing it could do without completely melting the system? That's what we're looking for.
- [19:40]
C
It's hard to define how big. I guess I could measure it in voxels. Is this is the best way. We're working on some new optimizations called sparse simulation that's going to make things even more efficient. But yeah, we've done some tests with our upcoming sparse solver for Embergen where we're able to do billions of voxels and it's kind of hard to describe in words, but if you think of a highly detailed, you know, group of fireballs that cover the entire screen and you can't see, you know, any artifacts or, you know, little squares.
- [20:20]
B
I want to make an earth, wind, fire joke. You have earth, wind, fire, water, four elements. You've got fire, you've got water. What about wind? Is that on the roadmap? Windergen, Winigen.
- [20:33]
C
It would be covered under Embergen. So yeah, we have people doing like snowstorm and fog and steam type effects with Embergen. But yeah, that's, that would be considered a, a gas.
- [20:46]
A
What about like rocks and dirt or something? Is that so?
- [20:49]
C
We have, we have a product called Geogen that does procedural landscapes and it's, it's pretty amazing as well. And then we also have a recently announced tool called Illugen and it's more of a procedural texture maker that's specifically focused for game effects artists. So yeah, we have a suite of tools just mainly talking about Amberen and Liquigen here because they're the most GPU compute intensive.
- [21:19]
B
I got a question. So on this slider thing, like does it go from no fire to the bomb Taza, like however you say it, the biggest nuclear explosion is that the slider? Like, I mean, where is the gaps here? Like where do you, where do you. Can you start on the slider? And then what is the biggest thing you can produce if you were just like. Because I want to do it, I want to max this thing out. Like, what is the biggest explosion we've done?
- [21:44]
C
Nukes.
- [21:45]
B
Okay, well, that, that. That's the biggest nuke that ever exploded, thanks to the USSR. I think it was 63.
- [21:52]
C
I think. Yeah. The SAR.
- [21:54]
B
Yeah, yeah, yeah. And the only fun, fun fact is they only ran it at 50% because they thought they were going to set the atmosphere on fire, which, you know, might have been true. Who knows? So it can go.
- [22:04]
C
Or they had plans for a bigger one and decided not to risk it.
- [22:08]
B
Well, I mean, probably smart, thank God, but so we really. It really can go. That is the scale from a match all the way up to a nuclear explosion, basically.
- [22:17]
C
Yes. Yep, Absolutely.
- [22:18]
A
So, Jason, how tightly integrated is USD support and Jenga effects tools? And how do you see USD shaping your future pipeline visions?
- [22:27]
C
So it's on our roadmap. We do get asked about it a lot. It is a. I'm not an expert, but my understanding is it's. It's a pretty tough format to tackle. So we haven't had the resources to do that yet. But yet we do get asked about it and we do have it in our plans.
- [22:45]
A
Very cool. I'm excited about that.
- [22:48]
B
So if you were to, you know, add a feature into Embergen that doesn't exist, assuming that you had unlimited dev resources and money and your CEO wasn't like, said no, what would be that one feature you would add into emergen that everyone asked for?
- [23:07]
C
Ah, this is really specific. But a lot of times we get asked about spline paths. So for instance, if you want to direct a fire tornado along a specific curve, right now, we can do that, but you kind of have to, like, I don't know, you kind of have to massage it a little bit with forces and things, but to be able to actually just draw the path. You want things to go for liquid Gen as well. And that's on our. That's on our road, dude.
- [23:40]
B
Fire tornado must happen. Let. Let's speak about this now. Here's another question.
- [23:45]
C
I can create one for you right here in like five minutes.
- [23:48]
B
I know. Well, I'm gonna. I'm gonna see it at seagraph anyways, so I'm gonna be. I'm gonna see how many nuclear explosions I could do before I crash the black wall. But no, here's a serious question. Embergen Liquigen, obviously all part of Jenga, but different tools. How do you handle something where there's fire and water together. Let's pretend a scene. Let's take. Not my favorite movie, but it reminded me of this is 300, not the original one, but the one with the ships. The one that wasn't quite as good. Remember the scene where they put all the oil on the water and they light it on fire? With the Greeks where you have water but you also have fire on top of the water and there's no separation. How do you handle that?
- [24:28]
C
So you would. You have to simulate those separately and then put them together in another tool. I have done that. I did a jet of water colliding with a jet of fire. And so what I did was I put a collider in the middle and so on the embryogen side it's hitting this invisible like kind of a noisy collider. And then the same thing in Liquigen. So but when they meet together and get all mushed up, you can't really tell that they're. They're not truly hitting each other in the same simulation. It's, you know, all VFX are smoke and mirrors. But there are tool, you know there things that Houdini can do that are like much more advanced but it's outside the scope of what we're trying to do where you, it's called, you know the, the technical terms for is like two way coupling of you know, where the force of the water will push a boat that will also push the fire and then the fire reacts to that. And so that, that gets into pretty complicated stuff. But.
- [25:42]
A
So Jason, have you seen any interesting use cases of emergen beyond just like you know, game dev and vfx? Are you doing anything in scientific visualization or AR VR yet?
- [25:53]
C
Oh yeah, I've seen all kinds of crazy things. I mean we don't, we do not recommend it being used for engineering applications because it's, it is designed specifically for you know, media and entertainment use. Now we do have scientific visualization where you're. It just needs to kind of get the idea across. But you wouldn't want to be making any engineering product design decisions based on like oh, this is flowing this way so we should change this. So there are, there are specialized tools that do that and as a matter of fact some of our liquid gen developers have a background in that industry as well. But as far as other fun stuff, like I mentioned the sphere, we saw that someone sent us something where they had done the visuals at the Oscars. I don't know if you watched it, but there was a Wicked duet thing and some of the background visuals were using Imperagen. So yeah, crazy stuff like that. You wouldn't think we had someone doing something for NASA like a lunar lander was coming down and kicking up the dust. So yeah, stuff you would never think of. There's some, I probably can't mention some customers names but there's, there's these big companies and they have these little pockets of R and D. They're just like these research centers where they're doing all kinds of exciting like 3D visualization. So those, those kind of calls are fun. Like you know, it might be military related or all these little industries that you don't really think about.
- [27:30]
B
That's cool.
- [27:31]
A
Very cool.
- [27:32]
B
Well, Jason, we're getting kind of up against it. Like what we like to do to close out the episode is to pretend that someone who hasn't listened all episode comes in right at this point because you know, recency effect. People remember what is at the beginning and what is at the end. Give the 30 second, one minute of what they need to take away about Jenga and Liquigen and Embergen and no pressure. Go.
- [27:57]
C
Yeah. So Jenga Effects, we make real time software tools for the visual effects industry. So it's used in games, films, advertising. Everything we build is highly optimized for the GPU and we really appreciate Dell's help in getting us the latest hardware and GPUs and everything to, to show it off. And then at Siggraph we'll be bringing all of our latest builds. Like I have experimental builds of things we're trying out. So our devs are excited to see what happens on the Blackwell stuff, see what the results are. And yeah, it just gets more and more fun to show off the more powerful these cards get.
- [28:36]
B
That's amazing.
- [28:37]
A
You guys certainly know how to draw a crowd. I mean it's true. Tons of people always around the demos. We're really excited.
- [28:45]
C
And I'm also very excited personally to get the new T2 with the full 600 watt.
- [28:54]
B
So Jason, do us a favor. Where can someone listening find and learn about Embergen and Liquigen, like what's the best website? And then obviously being a social media superstar like you are, where can they find you on LinkedIn?
- [29:07]
C
So just go to Jenga effects.com J-A N J G A F X.com and you can download any of our tools for free to try. So I'd encourage everybody to go do that. As far as me personally, I have a link trees you can go to jason key.com and it's got links to all my social media, including LinkedIn.
- [29:27]
B
All right, Cindy, any final thoughts before we close this beast down?
- [29:31]
A
No final thoughts on my end. Love software. Love the work that the guys over at Jenga FX are doing and really appreciative of the partnership.
- [29:39]
B
That's awesome. Well, in this episode we learn how you can make fake fire and water. I mean, that's basically the takeaway. And I mean, I'm very excited. We'll definitely, whenever we post this, get some videos to post in the comments where you can see when I try to make a bunch of explosions all at once and crash the black belt, which we'll see if that actually happens. So this is reshaping workflows with Delpro Max and Nvidia RTX GPUs. Until the next time, keep keep your Embergen and your Liquigen workflows running locally on RTX Pro and we'll see you on the next one.
- [30:31]
C
This podcast was produced in partnership with Amaze Media Labs.