Like others have said, it's interesting in theory, but this is definitely not an even-handed or straightforward pitch. The camera never stops moving in their shots (or even moves slowly) and never really zooms in on any particular detail. They do hardly anything to show the advantages that additional modeled geometry might have over modern normal-map plus poly modeling approach. The examples of the poly-modeling approach they show are all around five years old! To top it off, all their art (especially their logo) looks like crap. The whole thing comes across very unprofessionally and unconvincingly.
While I can conceptually recognize that this technology might have potential, this presentation is just awful.
Vaporware. Even if they do have some fancy voxel-what-ever-thingie, it has no use until you can do animation, proper lighting, shadows, transparency, post processing - none of which were present in any of their demo scenes or picture gallery? If those scenes were for real and a good demonstration of it's current features, it still seems to have a long way to go - and polygonbased videocards and tools keep on getting better...
All of the examples use a lot of repeating objects. I wonder if the system relies on a lot of instancing to run at those speeds, so that the search algorithm he was talking about can compartmentalize its searches to the point cloud data of one object and its position in space relative to the viewer rather than searching through a maelstrom of unique points that make up an entire scene.
This. Also it was all static images, there was no animation, only camera movement, what happens to those points once you start moving them around? whats keeping track of them then?
I for one, welcome our new voxel/point-cloud overlords, let the revolution begin . I've been wondering where all the cool voxel digging/building/terrain manipulation games are, ever since i saw those cool voxel engine demos by ken silverman: http://advsys.net/ken/voxlap/voxlap03.htm
Carmack, as far as I know, was opting for a hybrid solution since the polygon characters we have today already are looking very good.
Honestly if the general "shooting guys in the face" next generation looks like the character from nvidia´s human head demo This I would be freaked out.
This unlimited detail is probably a voxel octree thing.
I think voxels would be really usefull in making building plans for cars and stuff like that... maybe models of molecules/proteins etc.
but for games? the stuff the guy showed looked horrible :P.
A hybrid system would rock so hard, although i'm not sure it's even possible. These nifty points for the static stuff like environments and good ol polygons for characters
I personally do not care how bad something looked in quality for it being a demo. Anything that is mentioned about removing polygons as a possibility, or a thing to come is welcome to my eyes and ears any day.
You're all too negative and hard pressed to see only the bad in things...
I for one, welcome our new voxel/point-cloud overlords, let the revolution begin . I've been wondering where all the cool voxel digging/building/terrain manipulation games are, ever since i saw those cool voxel engine demos by ken silverman: http://advsys.net/ken/voxlap/voxlap03.htm
Have you played Voxelstein? It's awesome. You can take your little knife and chip away at a regular old stone wall until you make a hole through to the other side. You can also chip away at dead nazis and see their organs and bones and stuff. The grenades do the trick faster, but they're not very plentiful.
Skeptic, until they show me something as good or better than our current gen of graphics I'm completley unconvinced. All their art looked bad (including that logo grossssssss), and I don't care if you can have trillions of points if it looks like poop.
Until unlimited detail can have decent animation, decent models, great animation, good lighting and shaders, its a lost cause. Tessilation looks a lot more practical.
Until unlimited detail can have decent animation, decent models, great animation, good lighting and shaders, its a lost cause. Tessilation looks a lot more practical.
That's a bit short-sighted, isn't it. After all, we didn't start off with all of the tools we currently have for polygon modeling and animating, did we? It was a long process that took well over a decade of development to get to where we are now.
These guys aren't proposing that everyone should drop polygon modeling immediately. They are just proposing a different approach, and trying to drum up some investment capital to do it with. That is what their demo was for. They aren't selling this to end-user artists like yourself, they are trying to get funding to develop an SDK. The guy even says as much in the video.
They are starting off at the very bottom, and you are judging them by the standards of mature technology with over ten years and billions of dollars behind it. The point isn't to compare the two directly, but to show the potential that could be further expanded upon.
That's a bit short-sighted, isn't it. After all, we didn't start off with all of the tools we currently have for polygon modeling and animating, did we? It was a long process that took well over a decade of development to get to where we are now.
These guys aren't proposing that everyone should drop polygon modeling immediately. They are just proposing a different approach, and trying to drum up some investment capital to do it with. That is what their demo was for. They aren't selling this to end-user artists like yourself, they are trying to get funding to develop an SDK. The guy even says as much in the video.
They are starting off at the very bottom, and you are judging them by the standards of mature technology with over ten years and billions of dollars behind it. The point isn't to compare the two directly, but to show the potential that could be further expanded upon.
That's a fair post.
But to get that investment, they are going to have to come up with some demos that show the POTENTIAL of being better than today's graphics. Show off huge amounts of geometry, show off thousands of moving objects, demonstrate that fine texture detail is at least possible with the engine etc.
To get investors and industry insiders interested, they need better demos basically. As someone else posted, those demos don't look any more impressive than Outcast, using a 10 year old Voxel engine.
And seriously, what put me off more than anything was that the video seemed like a children's TV presenter explaining 3D graphics
The video should be aimed at insiders, as surely even potential investors are going to need some knowledge of the industry, and they need a much, much better website, with more info.
I understand your point, but the company seem too amateurish, and also the demos simply aren't impressive enough.
Infinite and unlimited do not actually mean exactly the same thing. You guys keep mixing them up, stop it!
One is utterly apart from finite numbers, eg it cannot be put to numbers, measurements, degrees -- the other is merely an infinite capacity for growth.
I am pretty sure if you look either one of those up in a thesaurus, they will refer to each-other.
Anyway, saying that a technology is unlimited is pretty damned stupid. Because in some way or another it will have a limit. If its good at instances, then a shiteload of unique objects will break it. If it excels by only rendering what is visible, you could always run a high enough resolution to make it die.
Thanks for the cool links. I'll be keeping an eye on this Atomontage engine.
Yeah, tesselation looks cool in all those uniengine demos, but to me it´s more an evolved version of normal/displacementmaps.
What you see in the Atomontage videos however, that really looks like a whole new language.
What I´d like is not for bricks to pop out more but rather a less restricted way of working.
I can't even begin to imagine how you would animate "points in space"
You realize polygons are already just "points in space", right? A bone tool could and should work the same for a dense cloud of points (voxels) as it does for a loose cloud of points (polygons).
Density is all we're really talking about here. Either we use voxels at the pixel level or we tesselate polygons down to the pixel level. Both techniques give us exactly the same end result: unique detail for every pixel on screen.
Horrible demo video for people like us, but it's not aimed at us. It's aimed at clueless business people who think of gaming as Pong and Space Invaders.
Seems an overall interresting tech but as people said what IF this stuff has to get into motion, how does it scale at for example a FullHD resolution etc. Even with regular polygons we can get a insane number nowadays but as soon light, shadows and other stuff add to it, we have a problem Houston.
I think it was Tekken4 on PS2 that said they could do models were so dense they didn't need textures. It was big deal then, but quickly brushed aside and never spoke of again.
I remember hearing Carmack's thoughts on octree voxel stuff a year or two ago, always thought it sounded neat. The "compare" video on the UnlimitedDetail site is a lot less condescending, though the narrator remains pompous - it's very internet-y, like he's making a thread on SomethingAwful or imitating Yahtzee. I'm just going to assume he's shy.
Paraphrasing: "the future for UnlimitedDetail is unknown; either we join with one of the major players (ATI, Intel, Nvidia), or not, in which case we're going to try and put this on weaker platforms in software mode, such as the DS, Wii, iPhone, etc."
... which could be neat, though I'm not really sold. Maybe it's just the bad presentation. Needs a lot of the things being mentioned here proven, too! But... for static environments, perhaps useful..?
I am pretty sure if you look either one of those up in a thesaurus, they will refer to each-other.
Synonyms do not actually always share the exact same definition. Just one thats similar enough to be interchangeable in certain instances. With the english language, this leaves us with a hell of a lot of synonyms that in many situations mean significantly different things.
god i was just being difficult and now you're making it into a whole thing. :poly127:
Anyway, saying that a technology is unlimited is pretty damned stupid. Because in some way or another it will have a limit. If its good at instances, then a shiteload of unique objects will break it. If it excels by only rendering what is visible, you could always run a high enough resolution to make it die.
Again, you have to think in terms of potential applications. What about using it for 3D GUIs and interfaces? If it performs as well as they claim, you could create some incredibly deatiled UIs using a system like that. As to the resolution, what about using it on consoles? Those usually only have a fixed number of resolutions that they can render, so there is an upper benchmark to shoot for. And what could this mean for cross-platform development. Software rendering solutions are much easier to port to other platforms.
Sure, this isn't a magic bullet. There are tons of problems and limitations with it. But there are ways around all of those. Do you think people had polygons all figured out at the beginning of that approach to 3D rendering? Of course not. They are still coming up with new ways to optimize and improve that approach.
It is easy to look down on this technology because it is different and unfamiliar. But that certainly doesn't mean it is an invalid or unwise approach. I wouldn't be surprised if several companies came forward at GDC this year with similar re-thinking of the traditional rendering pipe-line. It may very well be one of the big topics discussed.
You realize polygons are already just "points in space", right? A bone tool could and should work the same for a dense cloud of points (voxels) as it does for a loose cloud of points (polygons).
Density is all we're really talking about here. Either we use voxels at the pixel level or we tesselate polygons down to the pixel level. Both techniques give us exactly the same end result: unique detail for every pixel on screen.
Horrible demo video for people like us, but it's not aimed at us. It's aimed at clueless business people who think of gaming as Pong and Space Invaders.
Managing 6k tris is a lot easier than unlimited, but yeah, this is going to need a whole new toolset to even do basic things that we need with polygons, how are we going to model (3DCoat voxels?) texture and animate these things
Sounds good but what is the point of this? Theoretically this system could simulate a near perfect world, just like IRL. But then this would mean that the artist will be a bottleneck now, and only the best of the best will be able to make an asset that looks perfect.
Also, why bother trying to imitate the real world on a 2D screen? The monitor is still an impenetrable barrier that stands between the player and full immersion. If you want to be able to play the "IRL simulation", you should be investing money in brain research.
The monitor is just a limitation of current tech. We'll probably have overcome the detail barrier long before we break away from that into more immersive tech. Doesn't mean we shouldn't be striving for that goal till then.
ZD: any polygonal model or sculpt can be converted into point cloud, so creating it isn't really an issue; as for working with it, yes, I think it would require a new toolsettraditional rigging and animation programs certainly won't work on my computer if I tried to animate something of unscaling billion-poly density. Screenspace dot-cloud viewport renderer!?
Well what I was thinking was more of a "in your brain" simulation, much like a vivid dream, rather than projecting a rendered world onto your optics and stimulating senses or something like that.
So far the brain is the best simulator, although who knows, maybe when quantum computers become mainstream (if they ever, and they should) they could perform similarly.
maybe it's just me but the whole thing sounded like Peter Molyneux talking about how life-like the NPC's in fable 2 would be, or how much everyone would just instantly love their dog.
though this guy has given me a craving for some lettuce...
but who on god's green earth has the money for unlimited artists and designers?
If it was real it would take game devs generations (PS5 & xbox 1080) before we'd see companies using it to its full potential.
also the potential for overloading the senses(too much detail to process quickly) would increase dramatically.
I do agree that currently polygons are the best way to make out 3d games but I also think that it would be a shame to not explore other creative possibilities. Like Richard Kain was saying we havent even seen someone try to make something cool with this tech yet and its only in its first generation so its hard to make a judgement.
be aware it's a static scene and pre-processing takes pretty long
there has virtually been done no research on how animation would work with voxel representations, so I would be somewhat skeptical these "unlimited" guys have pulled that off from nowhere. I'd rather invest money in incremental improvements on established ways, that have greater benefit. (as in REYES like rendering, on-the-fly sub-division...).
Seems an overall interresting tech but as people said what IF this stuff has to get into motion, how does it scale at for example a FullHD resolution etc. Even with regular polygons we can get a insane number nowadays but as soon light, shadows and other stuff add to it, we have a problem Houston.
I had a look at the comparison video on their website, and to be fair, that has some more impressive demos. Particularly the foliage etc.
Regarding animation, well the way it would work shouldn't be too different to strapping some vertices to a constraint really. The logistics might be trickier, but the principle should be the same. They are still points in 3D space, the difference being, there are more points, and they are not bridged with polygons.
Collision detection on the other hand could be a bit tricker. Ironically I think you'd have to create some low poly geometry that was transformed but not rendered.
Even if you made a lower density version of the point cloud level, there would be far too many points to iterate through for collision, even if your nodes were as small as possible.
It would be more efficient to just have some standard polygonal geometry to find intersections etc.
I would like to know more about the mechanics of this though.
Funnily enough the sorting shouldn't be that different to polygons either, they probably have a depth buffer that works in a similar way.
Technically it's obviously very clever, to run real time in software, but like someone else said - to be honest I'm not a fan of the look of voxels/point cloud graphics. It looks too bitty and grainy, but not in a cool, pixel art way.
I prefer the look of polygons by far. Even now, Virtua Fighter 1 and Virtua Racing look good to me, I just love polys
That's one of the reasons why I'm not as excited or optimistic as some people here, like Richard etc.
I wouldn't be surprised to see this guy appear on Shark Tank.
It's exciting technology no doubt. Imagine a hybrid where this is used for the level and your GPU is free to work on characters and physics exclusively. Time will tell.
just give it a couple of years.. you never know. that guy says 16 months for release. i hardly believe anything comparable with what we see in our games now will be ready based on what he shows there BUT you never ever know.
pretty much that is how long such "experiments" last.
maybe if he asks one of the amazing artists here to help him some might change their mind and start saying how cool it is.
some people decades ago were saying gaming has no future.
Replies
While I can conceptually recognize that this technology might have potential, this presentation is just awful.
I remain skeptical.
This. Also it was all static images, there was no animation, only camera movement, what happens to those points once you start moving them around? whats keeping track of them then?
but for games? the stuff the guy showed looked horrible :P.
You're all too negative and hard pressed to see only the bad in things...
Have you played Voxelstein? It's awesome. You can take your little knife and chip away at a regular old stone wall until you make a hole through to the other side. You can also chip away at dead nazis and see their organs and bones and stuff. The grenades do the trick faster, but they're not very plentiful.
hahhaaha. no you can not.
I would say anyone here would welcome more detail and more visual fidelity, it's just that there's some issues with what was shown.
I got unlimited +1
But anyway
Until unlimited detail can have decent animation, decent models, great animation, good lighting and shaders, its a lost cause. Tessilation looks a lot more practical.
That's a bit short-sighted, isn't it. After all, we didn't start off with all of the tools we currently have for polygon modeling and animating, did we? It was a long process that took well over a decade of development to get to where we are now.
These guys aren't proposing that everyone should drop polygon modeling immediately. They are just proposing a different approach, and trying to drum up some investment capital to do it with. That is what their demo was for. They aren't selling this to end-user artists like yourself, they are trying to get funding to develop an SDK. The guy even says as much in the video.
They are starting off at the very bottom, and you are judging them by the standards of mature technology with over ten years and billions of dollars behind it. The point isn't to compare the two directly, but to show the potential that could be further expanded upon.
This seems like it'd be more useful in medical or simulation uses, and maybe for some offline rendering, until tools behind it get more advanced.
That's a fair post.
But to get that investment, they are going to have to come up with some demos that show the POTENTIAL of being better than today's graphics. Show off huge amounts of geometry, show off thousands of moving objects, demonstrate that fine texture detail is at least possible with the engine etc.
To get investors and industry insiders interested, they need better demos basically. As someone else posted, those demos don't look any more impressive than Outcast, using a 10 year old Voxel engine.
And seriously, what put me off more than anything was that the video seemed like a children's TV presenter explaining 3D graphics
The video should be aimed at insiders, as surely even potential investors are going to need some knowledge of the industry, and they need a much, much better website, with more info.
I understand your point, but the company seem too amateurish, and also the demos simply aren't impressive enough.
One is utterly apart from finite numbers, eg it cannot be put to numbers, measurements, degrees -- the other is merely an infinite capacity for growth.
Anyway, saying that a technology is unlimited is pretty damned stupid. Because in some way or another it will have a limit. If its good at instances, then a shiteload of unique objects will break it. If it excels by only rendering what is visible, you could always run a high enough resolution to make it die.
Yeah, tesselation looks cool in all those uniengine demos, but to me it´s more an evolved version of normal/displacementmaps.
What you see in the Atomontage videos however, that really looks like a whole new language.
What I´d like is not for bricks to pop out more but rather a less restricted way of working.
Density is all we're really talking about here. Either we use voxels at the pixel level or we tesselate polygons down to the pixel level. Both techniques give us exactly the same end result: unique detail for every pixel on screen.
Horrible demo video for people like us, but it's not aimed at us. It's aimed at clueless business people who think of gaming as Pong and Space Invaders.
Paraphrasing: "the future for UnlimitedDetail is unknown; either we join with one of the major players (ATI, Intel, Nvidia), or not, in which case we're going to try and put this on weaker platforms in software mode, such as the DS, Wii, iPhone, etc."
... which could be neat, though I'm not really sold. Maybe it's just the bad presentation. Needs a lot of the things being mentioned here proven, too! But... for static environments, perhaps useful..?
Synonyms do not actually always share the exact same definition. Just one thats similar enough to be interchangeable in certain instances. With the english language, this leaves us with a hell of a lot of synonyms that in many situations mean significantly different things.
god i was just being difficult and now you're making it into a whole thing. :poly127:
Again, you have to think in terms of potential applications. What about using it for 3D GUIs and interfaces? If it performs as well as they claim, you could create some incredibly deatiled UIs using a system like that. As to the resolution, what about using it on consoles? Those usually only have a fixed number of resolutions that they can render, so there is an upper benchmark to shoot for. And what could this mean for cross-platform development. Software rendering solutions are much easier to port to other platforms.
Sure, this isn't a magic bullet. There are tons of problems and limitations with it. But there are ways around all of those. Do you think people had polygons all figured out at the beginning of that approach to 3D rendering? Of course not. They are still coming up with new ways to optimize and improve that approach.
It is easy to look down on this technology because it is different and unfamiliar. But that certainly doesn't mean it is an invalid or unwise approach. I wouldn't be surprised if several companies came forward at GDC this year with similar re-thinking of the traditional rendering pipe-line. It may very well be one of the big topics discussed.
Managing 6k tris is a lot easier than unlimited, but yeah, this is going to need a whole new toolset to even do basic things that we need with polygons, how are we going to model (3DCoat voxels?) texture and animate these things
Also, why bother trying to imitate the real world on a 2D screen? The monitor is still an impenetrable barrier that stands between the player and full immersion. If you want to be able to play the "IRL simulation", you should be investing money in brain research.
My 2cents.
Slightly relevant things that come to mind:
blog.wolfire.com/2009/11/Triangle-mesh-voxelization
blog.wolfire.com/2009/11/volumetric-heat-diffusion-skinning
[ame]
So far the brain is the best simulator, although who knows, maybe when quantum computers become mainstream (if they ever, and they should) they could perform similarly.
though this guy has given me a craving for some lettuce...
If it was real it would take game devs generations (PS5 & xbox 1080) before we'd see companies using it to its full potential.
also the potential for overloading the senses(too much detail to process quickly) would increase dramatically.
i like my polygons. especially now that lighting in more realistic games has become much more pleasant to look at.
http://www.tml.tkk.fi/~samuli/publications/laine2010i3d_video.avi
http://www.nvidia.com/object/nvidia_research_pub_018.html
be aware it's a static scene and pre-processing takes pretty long
there has virtually been done no research on how animation would work with voxel representations, so I would be somewhat skeptical these "unlimited" guys have pulled that off from nowhere. I'd rather invest money in incremental improvements on established ways, that have greater benefit. (as in REYES like rendering, on-the-fly sub-division...).
some more
http://ompf.org/forum/viewtopic.php?f=3&p=17760
besides a not exactly professional looking website and logo, heavily looks like no money is involved, which doesn't add to credibility.
Houston, we have a problem you mean
Regarding animation, well the way it would work shouldn't be too different to strapping some vertices to a constraint really. The logistics might be trickier, but the principle should be the same. They are still points in 3D space, the difference being, there are more points, and they are not bridged with polygons.
Collision detection on the other hand could be a bit tricker. Ironically I think you'd have to create some low poly geometry that was transformed but not rendered.
Even if you made a lower density version of the point cloud level, there would be far too many points to iterate through for collision, even if your nodes were as small as possible.
It would be more efficient to just have some standard polygonal geometry to find intersections etc.
I would like to know more about the mechanics of this though.
Funnily enough the sorting shouldn't be that different to polygons either, they probably have a depth buffer that works in a similar way.
Technically it's obviously very clever, to run real time in software, but like someone else said - to be honest I'm not a fan of the look of voxels/point cloud graphics. It looks too bitty and grainy, but not in a cool, pixel art way.
I prefer the look of polygons by far. Even now, Virtua Fighter 1 and Virtua Racing look good to me, I just love polys
That's one of the reasons why I'm not as excited or optimistic as some people here, like Richard etc.
It's exciting technology no doubt. Imagine a hybrid where this is used for the level and your GPU is free to work on characters and physics exclusively. Time will tell.
pretty much that is how long such "experiments" last.
maybe if he asks one of the amazing artists here to help him some might change their mind and start saying how cool it is.
some people decades ago were saying gaming has no future.
Kind of a necropost, but I lulz'd.
But there is a difference between using it in a name metaphorically, and actually using it as a part of the sales pitch :P
No judgment being passed here, just saying it was funny :P