I don’t really know what’s going on here, but it is pretty nice looking. Isn’t this basically what voxel tech is?
Idk. But I want it.
I need it!!
another thread on “umlimited detail”? … the topic is that old and dead talked and it still finds people that are stunned by it although its basically game unusable 
We will see to what numbers they tone down their claims once they use it in a real environment that has collision, AI etc that needs power too, at the time its near completely rendering only in their tech demos and it seems to be years away from production use as I doubt anyone has a 200ghz+ supercluster under the desk to generate collision data on the fly for this amount of polys the representation of said objects would generate (cause rigidbody physics just needs it, alternative physics don’t do so but they in trade are magnitudes slower especially when you fire such amounts of data at them)
What about if for physics they don’t actually resort to atom based technology?
I mean even today’s polygon games, which contain physics, don’t apply rigid-bodies to every vertex of each model, right? So why would we assume they’re forced to apply physics to every one of those atoms in their unlimited detail meshes?
Because they have no option not to do it and only primitive colliders are useless for a game engine and the engine does not use any polygons to use a simplified collider as you do it with polygonal games (if it does, it defeats its own point as you have to generate both games then basically one in simpified for physics handling and one for the atom rendering).
Just so we’re on the same page, calculating the physics using polygons doesn’t require the polygons to be rendered, so if they had to bring in polygon to calculate physics, in what way does that defeats the point? Isn’t the point about having infinite detail?
But even so, converting stuff into polygons probably isn’t the only way to calculate physics, maybe they can figure out something that works better for their engine! (They’re the math experts! I’m just thinking out loud here).
The problem is, the engine has no triangles so also none to feed to the physics.
Its not like it does not render them but not having them at all.
Point cloud data are not generated out of polygon models (if you do, its meaningless as you can’t get more detail with point data if you didn’t have it already on the model) but generated like this right from the start (ETH Graphics lab has had a research modeller project available publically for some 5 years or so to model with point cloud data, rendering it with the point sprites. other applications like zbrush could learn it reasonably fast as they work with volumes not polygons), so you effectively need 2 types of assets to get polygon data for the physics and at worst have to model them in 2 forms or waste a lot of time optimizing it.
Thats why I don’t see the point in point cloud based rendering for realtime usage at the time, we don’t have the hardware power to operate with point based world simulations at the time on the larger scale.
I though see its use in “simple things” like ArchVIZ where the interaction with the environment is seriously limited and can at worst even be done by a recast alike meshification
the ‘atoms’ as he calls them can be generated from a model as he says in the video. You would only need highly detailed model to generate the ‘atoms’. You can use the highly detailed model to generate a much simpler polygon based model for physics to feed in the physics engine. You can do physics on the GPU since the CPU is handling the rendering. But I guess even if you can do the physics computations, moving the objects would be a problem, if I understand correctly.
I think it has potential, overall. Like you said, currently for other areas and maybe games in a few years as well.
The real issue isn’t whether this technology works, as there’s nothing in it that is considered technically impossible. (The “unlimited detail” schtick is marketing hype, but they’re hardly the first to fall into that particular trap. In theory, the technology they’re using is capable of handling unlimited detail, but you’ll need an infinitely large hard drive to take advantage of it!)
The crucial problem is how the hell do you make use of this technology?
The video suggests they have tools that can convert output from 3ds max and Maya, but the more polygons your artists can use, the bigger the datasets you have to handle, and the longer it’ll take to build your assets. A million-poly model isn’t going to happen quickly. Pixar’s animators take years to create their animated features, because it takes that long to model everything to that level of detail.
However…
The video shows no dynamic animations at all: it’s just a camera fly-through. Very pretty and detailed, but there’s no evidence that they’ve solved the physics integration. (It may be possible to handle physics at the “atom” level, but that would require some serious tools and coding.)
Collision detection shouldn’t be too difficult, but animating tree branches, grass, etc., is going to be an interesting engineering problem to solve as none of the existing off-the-shelf physics APIs will be even remotely compatible with this kind of engine. They’re going to have to roll their own animation, physics, collisions, and more. That’s probably going to take them at least another couple of years.
But it’s their constant bleating that this is all mostly “programmer art” and they don’t have a lot of artists that irritated me the most while watching the video. This rings more than a little hollow now. They’re in Australia, which is right next door to New Zealand—home of Weta Digital—and not far from Singapore and other popular offshoring art and animation centres. Furthermore, Australia does have a small games industry of its own, so there should be local talent available too, if out-sourcing isn’t viable.
In short: it’s interesting, and I don’t think it should be dismissed entirely. Polygons are not a panacea. But whether their engine lives up to the hype is another matter. I don’t think they’re doing themselves any favours by associating the word “infinite” with a piece of computer software: they know as well as anyone that computers don’t “do” infinity.
Yes you can generate it. But any model that you can generate a meaningfull model from that would work for this purpose can just as well be used directly in an engine, especially today with tesselation, in a streamlined and intuitive way compared to it.
It has definitely potential, but not if it tries to “sneak away from polygon tech users”, on that end its a lost cause as the workflows and the experience has matured for over a decade while point related stuff is in its child shoes as computers and technology just recently reached the level where it can reasonably consider going with it.
Out of my view it has to build upon its own benefits and innovate in its fields and mature especially. As others mentioned last year as well as this year, it still lacks even trivialities like lighting so its half a decade behind anything reasonable at all and thats no “light thing” to achieve.
True. As I know there is no way to animate things properly, especially a Character.
Nor any hope of deforming verts or anything like car damage. It’s amazingly limited and frankly, simply doesn’t look as good as any of the games the guy showed in that vid.
What he’s hoping for is for YOU to envision the possibilities. Sadly it doesn’t get better.
The low poly tree you saw in that game, overall looks better and serves its purpose, and can easily be upgraded to be ‘unlimited’ polys with a decent tesselation path.
It’s obvious to me that this is the future of rendering in 3D, maybe 5 or 10 years out though. At this moment it may not be ready but the whole foundation for 3D is to mimic reality and this is the next logical step, since indeed there is nothing resembling “polygons” in the makeup of reality. Triangles are just a primitive mathematical method to solve the problem and it works pretty well but we must eventually let it go and use a more “accurate” solution and it will be interesting to see it’s further progress.
Maybe these guys pre-rendered those videos, or they could be running it with a super computer, I’m not saying that can’t happen.
But personally I would wait a lil bit before saying this is only a scam, or even useless. I mean they made a camera fly from a 1 km high angle, down to a grain of dirt, all with maximum detail. I don’t think that’s just the old Comanche game extrapolated to today’s rendering speed, seems to be more capable than that.
I’m sure if what they claim happens to be true, they can also do something about animations and physics. They’re already claiming to be able to deal with the AO and shadows. Just because we can’t get our head around something, doesn’t mean they can’t do it either! I know I’m no mathematician!
It’s all about waiting and see!
actually you don’t need a super computer.
With the cut of interaction and other intense things I am pretty sure that a regular buyable machine today is able to handle it. With up to 4GB VRAM and high end cards like the 2500USD+ NVIDIA CAD cards and tesla you have more power under your desk than the average user can even imagine in his wildest dreams 
i have a great idea! put over 1 million extrudes on your model then each vert will be 1 atom! :]
Thing is we said all these things a year and a half ago. People said no animation, no decent lighting or effects, no physics, and where’s the demo with decent none repeated quality? They came back 16 months later and said look! bits of mud! And remember… polygons are rubbish. Now, this is pitched as a game tech. If you had a year to improve something really really detailed, would you spend that year scanning a rock or doing something about the fact things can’t move or collide?
They not only haven’t got around to implementing these things yet, but in the main they have scrupulously avoided even acknowledging them. Brash claims, no substance to back it up, and an unwillingness to respond to informed questions - does that combination usually fill you with confidence?
It’s not the future of rendering, its a dead end.
I don’t see any use for it other than stationary objects, like terrain.
They’ve never shown anything worth a second look. They just expect you to imagine what things you could do if you give them money.
And Notch from Minecraft goes as far as calling it a scam by snake oil salesmen.
http://notch.tumblr.com/post/8386977075/its-a-scam