⋮    ⋮  

Atomontage Inc.’s Branislav Siles on the Limits of Polygons, the Voxel Future, Streaming, AI and more


Polygons have their limitations. Just look at the leaps between console generations and you have a fine example of that. Each generation is capable of rendering more polygons and more details, but the diminishing returns are clear. Accidentally clip the camera behind some scenery in any given game, and you quickly learn that the game worlds you explore are very hollow. Only in a game like Minecraft are you allowed you really dig through and interact with the world, only to discover that there is far more world available to you behind it.

That is the potential of voxels. Now imagine that, on a tiny scale. That is the prospect that Atomontage Inc. is bringing to the table. Fully realized worlds that are solid, and can be dug into in real time, while retaining much of the detail of high poly meshes. Sounds too good to be true, but after talking with Atomontage's Branislav Siles, I'm almost a believer.

Euclideon’s Infinite Detail Engine Is Getting Two Titles – Is the Future of Gaming Here?

While at Reboot Develop Blue 2019 in Dubrovnik, Croatia, Siles sat down with me and showed me numerous slides, videos and examples of what his team is working on. Branislav himself has been working on the technology for years and founded the company with his partner initially to realize the game he had in mind.

Branislav Siles: Long story short, I wanted to make a game, a different one than all the other games. So I knew that I have to create technology for this game. And it has to be different than the others, so I decided to use voxel technology because I had some previous experience. Voxels, basically, are like Lego bricks, they fill the objects, so objects are volumetric.

BS: So these alterations happen in real time, because it's optimized, it's kind of cheap. This was running going on a computer, a gaming PC, from 2004 or 2005. So we've been developing this technology on laptops, which were kind of outdated, but it's okay, because if you develop the database, which is the key component, the key part of the software, if you develop it with a physics simulation in mind, you end up with something different than if you develop this stuff with just rendering in mind. That's what most people need, and they end up being almost useless technologies that could render something that didn't allow the user to edit geometry. Because the database is basically very different. Physics simulation is so hard to do, that if you make that, then everything else will be easier to do.

BS: In 2012 I demonstrated the technology and I made some public demos. Many companies were interested, like Crytek and Activision Blizzard, they basically wanted to buy something. I can't talk for them, but it looked like some of these companies want to buy this project, but I have my plans of course, and never said yes. So just went on developing the tech further until it became commercially viable, so to speak.

Apple Watch Shipments Reached 11.8 million in Q3 2020

BS: And then we set up this company, Atomontage Inc., and our first investors, Tommy Palm, who was the guy behind Candy Crush Saga, and then his father who is a medical specialist, and we started to set up a company in the US and in Slovakia, that's our subsidiary, fully owned by Atomontage Inc. And we are setting up a team of developers in Slovakia.

Branislav has a laptop which he uses to pull up slides and videos of Atomontage's technology. He explains to me the possibilities, such as generating voxel models on the fly and freely editing them. And then, before I know it, he shows me the technology running in real-time, on the laptop. And make no mistake, this was not a high-end gaming laptop, this was a pretty standard machine. And yet, the technology ran smoothly and performed just as in their videos.

BS: So what I tried to do is create this ordinary geometry, it hasn't been edited by people, it's basically mathematical functions, fractals, with some textures used as the source information for the generators. And as we see this is very complex geometry, it's not just the surface, it's real, very complex shapes. And if you wanted to use polygons for this, you'd end up using, I guess, hundreds of millions of polygons just to make the whole thing work. But the great thing about this is once it's in the form of voxel models, you can edit the geometry and simulate it destructively. So you have all kinds of freedom here in interacting with the content and world. Like in Minecraft, for example.

Branislav tells me about Atomontage's vision of a large world where players can edit the geometry online on the fly to create new things. It sounds promising, but much like Minecraft, you can easily imagine the world quickly becoming full of holes, or filled with phallic monuments.

BS: Part of our vision is that, at some later point, when our partners develop actual games, millions of users will have complete freedom over the content. Because it's so easy to edit, you don't have to learn extremely complex tools. Today, the cost of game development is extremely high. And it's going up all the time.

BS: There are good reasons, and there are bad reasons for the cost. So the good reason is that the quality is going up. But the bad reasons include that polygons are no match to the real world, right? If we want to do it right, we need volumetric content. So we can do simulation and interaction the way it's done in the real world, then it becomes cheaper, and accessible to all kinds of people, not just extremely skilled people, or those who can afford expensive tools.

BS: We cannot judge the whole industry based on Minecraft. But Minecraft has shown that you can have 10s of millions of children who don't have to learn complex tools, then they will make it the biggest hit in history. So what we really need is something like that, but without the limitations of Minecraft, and that's where powerful voxel technology won't just help, it'll be essential. Once that technology is good enough, it will allow you to not just create geometry cheaply by hand, but also generate geometry and physics simulation, where biological processes and chemical reactions can alter the geometry on the fly, so you end up with a fully interactive, living environment for millions of people. That is interesting because we have complex processes going on. And most of the alterations to the geometry in such a large space will be done by the simulation and logic processes, not really by people. And that's important because it won't become a boring place, it will be ever changing. And it will be believable because we're kind of used to those environments where things work and change over time. That's how the real world works. So I would say that's a key property of such an environment. So yeah, there are technical issues, but we have solved them, so to speak. And we're showing our demos that prove that our concept, our approach, is the right one.

BS: The latest streaming projects from Google and others show clearly that streaming is the way to go. But at the same time, you want to stream to millions of devices, those devices are very different. You want to stream dynamic content that's ever-changing. So very high poly meshes, with very high-resolution maps, it's not really the way to go. What we have is a streamable voxel format that explores the potential of levels of details.

I found his comments on cloud streaming interesting, so I had to ask whether or not the voxel technology would make game streaming easier.

BS: Yeah, much easier, because, first of all, you remove the latencies. Normally it's complicated, you get input from the user, where he is, how his head is oriented in space, tell the server, where it renders the picture, and then it has to encode it and send it to the client, which has to render it. So it's multiple sources of latency and amounts to, like, 50 milliseconds, it can be maybe 20, in a really great case, but it will more likely be higher if you don't have the best connection ever. But if you have volumetric geometry, whatever geometry you receive from the server, the client can render it instantly, because it's easy to render. The important thing here is that if you're in VR, when you turn your head, you're the last version of geometry you have, but you're rendering it now, so we don't have the latency of all those processes. So you actually can do VR with this on the cloud, and this is pretty important because everybody will jump into VR or AR at some point, right?

BS: So you can't have those latencies, because everybody would throw up, right? But when you have the volumetric geometry on your client, you just render that geometry instantly. And you can do that in VR. So it's not making something better, it's enabling the technology for that kind of scenario. Without it, you will have those latencies. But of course, from our perspective, the typical gaming applications are close to our hearts. So, yeah, these are the destructible vehicles...

BS: Media Molecule's Dreams is a very nice technology. But we haven't seen realistic content or something that could become realistic anytime soon. There's also Claybook, it's heavily physics simulation driven. Again, a nice example but very limited to a certain volume in space. Our technology doesn't have these limitations, and our technology provides all those key elements like voxelization, rendering of large environments, editing in real time, physics simulations, streaming. So it provides the key features that will be used in those big projects we can see coming in the near future.

Although the technology is coming along well and Branislav was keen to show me what he had, he also had to make clear that some of the more recent demos developed by Atomontage I'm not allowed to see, though they will eventually be made public.

BS: We cannot show you our Atomontage demos that are under NDA, but sooner or later, we will show them publicly.

BS: So it will be ready to follow our project in the next few months, we will also release some new examples we have. Because apart from doing marketing, we are in touch with the partners who want our technology. So what we are trying to do with our online presence is show the talent out there that we exist, and that they can work on this kind of technology, they can join us. In the next couple of months, we will release some more examples, some better geometry, to attract this talent, from mostly Europe. In about a year, we will set up additional offices in Stockholm where my co-founder is based. And later on somewhere else, we'll see how we will grow. We're now growing at the rate of about one person a month. Which is, I would say, our limit, we cannot grow faster. And it doesn't make much sense.

At this point, I'm pretty much sold on the concept of voxel technology being used in the kind of games I already play and love, but needed clarification on one point. Was this a whole new game engine, or middleware?

BS: Middleware I would say, Atomontage cannot really compete with Unreal Engine or Unity. It is possible our resolution could, at some point, become part of one of those ecosystems, who knows. But for now, it's very clear to us that what we have to do is cooperate with very specific partners, like the big guys we are working with already, who have longterm projects which are very interesting for us. Because we cannot work with hundreds of developers, we can work with one or two, as long as we are very small (we're six and a half people now at Atomontage). So we cannot work with more than two or three players. So for the next year, year and a half, we will focus on these cooperations. And later on as our SDK we're developing right now becomes more powerful, more robust, and more easy to use, we will provide it to more game developers also. But we are not game-specific technology, of course. So we can help companies in pretty much all industries where they deal with 3D graphics.

BS: Maybe in medical, but we see opportunities in AI also. Because if you think about robots, cars, whatever, navigating a space that's changing, it's very useful to have a technology that can provide the scale, the detail, but also the possibility of updating the geometry or changing the geometry and the real world at the same time. So it's AI, it's geographical systems, games, streaming. Anything, basically.

I came away from my chat with Branislav somewhat bewildered but equally fascinated. On the one hand, I'm convinced this is the future of gaming. On the other hand, I'm not sure it'll be quite the revolution that Atomontage hope. To see more of Atomontage's technology, and perhaps secure yourself a job, head over to their website.