I don't think it's ever going to out. This guy does the same thing every few years.
Release a video of his tech demo, talk a bit how 'polygon companies' are evil and banding together to make games look flat.
What ever happened to the sdk he was going to release last time? I'm not saying that a voxel based graphics could be the future. I just can't see this company being the ones to do it.
Its just the same marketing spiel they've always been giving. Just with newer art and no sign of release.
2003:
I'm betting that they promised that they where just *months* away from getting there engine to market back then too.
Im with Snefer and Spudnik on this one. We artist should keep an open mind, we are visionaries for christ sake. As I see this, its a totally new thing that people fail to see the use in, if it shows with time that this is not suitable for games there are alot of areas this can be used in. But like all new things that sound too good to be true (and the austrailian guy sure as hell does try to oversell it) do get met with sceptisism, but i do say that it astonished me to see that not more than 3 or 4 people actuallty saw some potential in this :S
Basically this.
We're supposed to be some of the most open-minded folk in any industry, yet new technologies get this much flak?
Sure, they may not have all the answers yet. As a society, we know so little about the universe so that is a given.... which is why I always hit my head against a wall when people willingly CHOOSE to be close-minded to new ways of doing things.
Change is how we became what we are today. What would have happened if as a society, the development of the computer was just ignored and ostracised? They'd still be trying to market the technology now to a select group of people.
I just think it is funny that they come back, a year later and show MASSIVE improvements over what they first showed.... yet they still receive the EXACT same flak.
Why hasn't this technology taken the industry by storm?
No, it has little to do with physics, animation, any of that. That can be worked around because we are humans. We find solutions to EVERYTHING. We don't stop just because 'aww it didn't work', we PUSH ahead and that is how we are what we are today.
The issue is that people are simply not comfortable with change as a whole. Now with the polygon conversion tools, that is closer to becoming moot. As well as the fact that point clouds are just geometry; point-cloud rendering systems can be put into any rendering engine. Additionally, you can paint directly on the model or unwrap, and that texture data will also be transferred to the point-cloud models as seen in their demonstration.
They also showed animated point clouds in their last video. Seems as if everybody wants to just make an instant decision on this technology without actually doing any research on the subject.
Don't care how bad it looks. That is point clouds being animated. So, major downside for point clouds #1 is no longer relevant, as it never existed as a concern to begin with.
Physics: Atomonage Engine has already displayed that.
Here is the true potential in regards to the graphical fidelity of point clouds.
19 seconds onward. Sure, it may be raytraced, but it can just as easily be rasterized and extra shading power could be dedicated on the GPU since the geometry can be rendered entirely on the CPU... on one core at that.
Given the current direction that most art assets are going in, I would think this kind of technology would just be a matter of time.
Every game these days seems to use normal maps that reference extremely high-detail models. Digital sculpting is swiftly becoming the de-facto method for adding detail to game worlds. Almost everything is starting off as high-detail digital sculptures.
A tech demo like this one is just showing someone's efforts to bring full-on digital sculpting into the rendering space. I don't see any reason why this couldn't be used as a foundation for modern game rendering. Sure, there are features that need to be addressed. But the same was true of how 3D game rendering was originally handled. This is simply the first step in a different direction.
Some people may remember a certain video on youtube a while back. It was a presentation of new technology that would allow game developers the freedom of unlimited detail when making 3D assets for games, throwing polygons away for a point-cloud system.
I have very little knowledge on the subject, so I don't really have a real opinion. It certainly looks too good to be true. One question I have is how will they handle animation?
Anyways, what do you guys think?
I still remain skeptical. Too many unanswered questions. How are they dealing with materials and shaders, lighting, animation. Changes in environments through destruction or something as simple as a door. How does this work with modern 3D hardware being so geared towards rendering triangles.
Clearly, something like this will one day happen. Maybe this is it, maybe not. It will take more than whitepapers and recorded visualizations to prove to me that this is more than a money trap for VC.
But more importantly, we still have a long way to go with tools to fully utilize even this generation's ability to render a comparibly limited amount of polygons. Projects are still struggling with constructing an environment on schedule and in budget within the current technology. Throw 'unlimited detail' in the mix, and you're talking about much higher than film level of quality environments. Beyond what's needed for the shot; a fully interactive 3D environment.
Sure, there's some tradeoff with not having to jump through the extra hoops of projecting high-poly models on to low-poly assets...but that doesn't make up for having to model invidual pieces of sand and blades of grass to populate your worlds. A lot of these tricks for dealing with limited polygons in a game engine are also incredible time-savers.
When 3D was first created, it was not created for making games.
Unwrapping for textures, bones for animating, collisions, physics... we had to come up with all of that ourselves to make 3D viable for games.
The same thing simply needs to be done here. We've shown that as humans, we can pretty much solve any problem.
Also, remember that the scenes shown are being rendered on a single-core CPU... completely in software rendering. Think about how much processing power that leaves developers with today. Even on a console such as the Wii.
I'm not quite convinced yet, but if there was some gameplay with this technology in the later future maybe I would reconsider it. For now I can't make a judgement of a fly through.
We're supposed to be some of the most open-minded folk in any industry, yet new technologies get this much flak?
Sure, they may not have all the answers yet. As a society, we know so little about the universe so that is a given.... which is why I always hit my head against a wall when people willingly CHOOSE to be close-minded to new ways of doing things.
Change is how we became what we are today. What would have happened if as a society, the development of the computer was just ignored and ostracised? They'd still be trying to market the technology now to a select group of people.
I just think it is funny that they come back, a year later and show MASSIVE improvements over what they first showed.... yet they still receive the EXACT same flak.
Why hasn't this technology taken the industry by storm?
No, it has little to do with physics, animation, any of that. That can be worked around because we are humans. We find solutions to EVERYTHING. We don't stop just because 'aww it didn't work', we PUSH ahead and that is how we are what we are today.
The issue is that people are simply not comfortable with change as a whole. Now with the polygon conversion tools, that is closer to becoming moot. As well as the fact that point clouds are just geometry; point-cloud rendering systems can be put into any rendering engine. Additionally, you can paint directly on the model or unwrap, and that texture data will also be transferred to the point-cloud models as seen in their demonstration.
They also showed animated point clouds in their last video. Seems as if everybody wants to just make an instant decision on this technology without actually doing any research on the subject.
Don't care how bad it looks. That is point clouds being animated. So, major downside for point clouds #1 is no longer relevant, as it never existed as a concern to begin with.
Physics: Atomonage Engine has already displayed that.
Here is the true potential in regards to the graphical fidelity of point clouds.
19 seconds onward. Sure, it may be raytraced, but it can just as easily be rasterized and extra shading power could be dedicated on the GPU since the geometry can be rendered entirely on the CPU... on one core at that.
This diatribe is so full of assumptions that one would suspect you have a personal, vested interest in this company and their technology being taken seriously and at face value.
Big difference between skepticism and close-mindedness.
I'm not quite convinced yet, but if there was some gameplay with this technology in the later future maybe I would reconsider it. For now I can't make a judgement of a fly through.
....like your consideration matters to the future of unlimited detail? Please....don't flatter yourself.
I think alot of us are maybe worried we wont have a Job anymore lol.
J/K
I agree with one of the comments, lets see it handle animation, physics, and collision and still be playable real-time
What? Why? Last i checked it wasn't called free detail, you still need artists.
There's this growing consensus that you must achieve the same level of detail as what was seen in the unlimited detail tech demo videos.....for goodness sakes - its just a tech demo video...not a "YOU CAN'T MAKE GAMES WITH HAVE LESS DETAIL THAN THIS" video.
The less whimpering and more doing happens the smoother and faster the transition will occur and a resultant win-win for all.
This diatribe is so full of assumptions that one would suspect you have a personal, vested interest in this company and their technology being taken seriously and at face value.
Or that maybe I am open to change, or have an open mind, or as an artist, I can appreciate the changes it would bring?
Big difference between skepticism and close-mindedness.
There is a very small difference between skepticism and having a closed mind. They both lead one to stick to their ways and do not allow one to grow, change, and adapt.
I say that because a true skeptic will search out the information... but most people who say they are 'skeptical' are saying that because they are closing their mind to the idea, and refuse to learn anything more on it.
Trying to attack the individual because you refuse to acknowledge the actual information.... classy. :thumbup:
It's not strictly a matter of open or closed mindedness. I think many people here agree that the tech is very cool and opens a lot of possibilities. But it's also a huge investment to make. It's new, and probably won't be an easy module to integrate with existing engines. Once it does, the entire studio would need to be trained on using it in and out. Limitations, learning what works, what breaks, and pushing a game into it? That's a serious investment of time for a studio.
Euclidion will really need to push hard to get someone to make a cool game and get this working on existing hardware for anyone to take a serious look.
Also, point B which everyone is already talking about is the presentation. He's making big claims using a style I generally associate with liars and mentally unstable people. I think pretty much everyone has picked up on that to some degree. Makes this hard to believe in, but I do believe they (him?) do have an actual product.
But, his presentation style is certainly effective in getting the name/idea out there and talked about.
Show me lighting, animation, shaders, physics, etc. and then I'll jump on board. There is no denying that the "geometry" part of the demo is intriguing, but it is still so far off from being a useable technology in game development I have no idea why they keep showing it.
Also the guy talking pisses me off.
And unlimited detail doesn't always make a game look good.
Infinitely 'instanced' more like! I smell some really shitty optimising going on in the background. I'd like to see a totally *unique* environment as a selling point, not something composed of insane amounts of instanced assets all over the place.. Isn't the whole premise behind this type of 'voxel'(esque) tech to allow "uniqueness at a level never before seen"?
Right now the animations are the mesh stored for each frame of the animation. Essentially the same as vertex animation. This has a major drawback, which is no dynamic animation, so no ragdolls, no wind making trees sway, or physics influenced animation.
Memory is the other issue with the animation setup, since you're storing each frame of the animation, you have to store the mesh for each frame. Being a point cloud, that will eat up a decent amount of memory per frame. Now multiply that by say even just 30 animations and you have quite a large file. Toss in the fact you'll need more than 1 character on the screen and forget about dynamically loading an animated point cloud into memory on the fly isn't really feasible due to the amount of time required to load from the hard drive (getting better with ssd's, but most people don't have them still).
Trying to attack the individual because you refuse to acknowledge the actual information.... classy. :thumbup:
What information? There simply is none. There are no live demos, no patents, no sdks, no technical reports, nothing to even imply that the product actually works. So again, what information are you refering to?
Right now the animations are the mesh stored for each frame of the animation. Essentially the same as vertex animation. This has a major drawback, which is no dynamic animation, so no ragdolls, no wind making trees sway, or physics influenced animation.
Memory is the other issue with the animation setup, since you're storing each frame of the animation, you have to store the mesh for each frame. Being a point cloud, that will eat up a decent amount of memory per frame. Now multiply that by say even just 30 animations and you have quite a large file. Toss in the fact you'll need more than 1 character on the screen and forget about dynamically loading an animated point cloud into memory on the fly isn't really feasible due to the amount of time required to load from the hard drive (getting better with ssd's, but most people don't have them still).
What if you use traditional skinned and animated polygon geometry that serves as a cage driving point-cloud data deformation?
We're supposed to be some of the most open-minded folk in any industry, yet new technologies get this much flak?
Sure, they may not have all the answers yet. As a society, we know so little about the universe so that is a given.... which is why I always hit my head against a wall when people willingly CHOOSE to be close-minded to new ways of doing things.
This is not some new alien tech, others have done similar things, people working in the industry has touched on this, even Carmack who is one of the visionaries you speak of, even he is man enough to admit the limitations of mega-texture, where it shines and where it doesn't.
The issue is that people are simply not comfortable with change as a whole. Now with the polygon conversion tools, that is closer to becoming moot. As well as the fact that point clouds are just geometry; point-cloud rendering systems can be put into any rendering engine. Additionally, you can paint directly on the model or unwrap, and that texture data will also be transferred to the point-cloud models as seen in their demonstration.
Point cloud data is not geometry, you cannot simply put it into a rendering engine and get geometry from it.
They also showed animated point clouds in their last video. Seems as if everybody wants to just make an instant decision on this technology without actually doing any research on the subject.
I would agree with this, not in the way you think though.
Don't care how bad it looks. That is point clouds being animated. So, major downside for point clouds #1 is no longer relevant, as it never existed as a concern to begin with.
As someone mentioned, animated, not skinned.
One of the downsides with this is that the geometry for each object has to be stored somewhere, the unlimited detail isn't generated from thin air, these are massive objects. mesh-per frame animation would just be silly in this scenario.
It just works magic by instancing the same meshes.
Here is the true potential in regards to the graphical fidelity of point clouds.
19 seconds onward. Sure, it may be raytraced, but it can just as easily be rasterized and extra shading power could be dedicated on the GPU since the geometry can be rendered entirely on the CPU... on one core at that.
No, that's exactly the same use we've seen of point cloud data before: a huge amount of static point cloud data and a smart way to read it, you don't simply just share work in the computer, you have to take into consideration how it actually does things.
I had a quick chat with one of our programmers on this tech and he got scared by the amount of data that will be required making this a usable technology. He thinks that's the biggest bottleneck, for it to be implementable for video games. Blue-rays aren't large enough, you would need something way bigger, closer to the direction of a terabyte, or more.
I dont see why folks are thinking physics would be such a problem. you would imagine that rigid body physics would be approached the same as in current generation games. that is assuming that objects can be moved efficiently in object space.
The pie is split it up for geometry, textures, animation, physics simulations, dynamic lighting, character ai, audio effects, etc. Things are optimized for a reason, it's not because we all love polygon limits and compressed lossy audio.
Game engines are a series of compromises. You could put a Zbrush model into UDK and talk about how amazing it looks, but it's going to consume your budget. You could use 2048x2048 maps on everything, but it's going to impact your memory budget.
These guys are shortsighted and arrogant. Don't compare your rendering engine to fully featured games, and then claim how ugly and stupid they are.
The pie is split it up for geometry, textures, animation, physics simulations, dynamic lighting, character ai, audio effects, etc. Things are optimized for a reason, it's not because we all love polygon limits and compressed lossy audio.
Game engines are a series of compromises. You could put a Zbrush model into UDK and talk about how amazing it looks, but it's going to consume your budget. You could use 2048x2048 maps on everything, but it's going to impact your memory budget.
These guys are shortsighted and arrogant. Don't compare your rendering engine to fully featured games, and then claim how ugly and stupid they are.
Exactly! Well said.
And about the animations, skeletal deformation is still going to be preferable to point cloud animation. Animated point cloud data is pretty much what Quake used and is the same as per vertex animation? But instead of drawing faces between the points and using less points you use so many points that its so dense you can't tell there aren't any faces between the points?
There is a point when it becomes so dense and so much data is floating around you're better off using less points and drawing faces between them.
It seems kind of weird that anyone would think that a half finished system that strips out the faces, packs the dead space with points and provides less tools and less functionality would get any kind of positive feedback.
Artists are visionaries yes and we 3D monkeys embrace new and weird stuff all the time but part of being a smart monkey is being able to discern what is useful from what is pointless and counter productive.
Edit: Apparently, its a scam I leave my post though...
************************
Well, from what they show, it looks very impressive.
However, everything is static and appears to be colored by the little "atoms" themsevles. Looks like there aren't any textures involved, I could be mistaken though.
It certainly is an interesting new technology and I'm eager to see how this turns out.
However, I noticed how they/he talks about the rest of the industry and years of improvement like total garbage. I love the part where they show footage from Crysis 2 and tell us that their technology could make it look thousands times as good
Anyone noticed how he points at backgroundobjects and tells us that these are just fake cardboard meshes. What does he expect artists to do? Spending weeks on a backgroundobject that will be seen like 5 seconds in the game?
Artists get payed and they are not cheap, studios have releasedates and publishers no unlimited budget to invest unlimited time in a game with unlimited details.
How is lighting supposed to work with this? Shadereffects? Any possibilitie to texture these models? Animation, dynamic objects? (Notice how absolutely everything is static in their preview scenes)
Well, I'm no professional and might not be up to the latest knowledge and tech'.
But this looks a bit hard to believe to me. Not the technology itself makes me doubt it, but the presentation.
What if you use traditional skinned and animated polygon geometry that serves as a cage driving point-cloud data deformation?
Can't see that as being efficient. This particular tech is more akin to a search algorithm...which doesn't lend itself to dynamic data. This is data that is being managed and organized in very particular ways beforehand to make the method of finding the right points to display as quick as possible.
Certainly, an engine could use some sort of hybrid approach. Building the static environment out of point-cloud and any dynamic geometry could use polygons. But here's the thing...at some point, throwing more triangles at an object really doesn't do any good. When the triangles are smaller than pixels, there's no point in rendering more triangles. If they're off-screen, there's no point in rendering them. If they're occluded, there's no point in rendering them. The end goal of a visual media is not to recreate reality, it's to create a convincing illusion of reality. Even Hollywood doesn't work with unlimited detail, and they've got one of the best rendering engines every created.
Right now I believe that by the this type of tech is truly ready for production the traditional polygon rendering engines will either have the raw hardware horsepower or developed enough new tricks to be able to render as much detail as you could actually need. I could very well be wrong about that, but with almost the entire the industry devoted to improving the current method over throwing everything out and reinventing the wheel, I'd be very surprised if polygon rendering was replaced anytime soon.
Either way, it really doesn't matter to the artist. What bottlenecks art in games now is generally not an inability to throw more detail on the screen. It's how long it takes to create those high-poly assets and the hoops one has to jump through to get those assets into the current engines. I feel we've already hit a bit of a ceiling on this...that modern engines are more than capable of rendering detail that most companies simply don't have the time, manpower, or technical proficiency to fully utilize. So it essentially boils down to the tools for creating these assets and the pipeline for getting an asset into a game.
People can get all tingly about how exactly about how unlimited visual detail is rendered, but ultimately it's how all that unlimited detail gets created that needs to change first.
how so? he stated the opposition side of the advocates fallacies.That entire tech demo is nothing but a huge fallacy. Its good that he is doing this but and playing up his work and causing a big stir. But he could have done it abit better while stating the pros and cons so far.
And noted the projects that should be getting the attention, projects that don't pretend do be something they're not.
Interesting but would be 10 times more interesting if they actually discussed the engine and their new system without dumbing everything down and not using good examples of how well it actually handles these enviroments in real game issues such as fps, dynamic lights etc.
I knew my common sense was tingling for a reason, mostly because if this thing were real they'd have been bought by Autodesk or Pixologic or somebody who has an interest in pushing millions of voxel-like "atoms".
It's obviously fake, if anyone still doubts, just do the math.
They say they use 64 'molecules' per mm3, That would mean
they use 64,000,000 for a single square meter. <br><br>One 'molecule' at least
uses integer-datatypes as positions for x y and z, which would be 4 or 8
bytes each, then they would need something to discribe the color,
lightning and the reflection of the molecule. I guess about 50 bytes per
molecule.<br><br>50x64 000 000 = 3,200,000,000 bytes of data per frame per
m3.<br><br>Say you have a room of 3x4x6 meters, and you want
to render it at 60 fps. Let's do the math...
3200000000x3x4x6x60 = 13,824,000,000,000 bytes per second to send
through your computer.
<br><br>That equals 13500000000 KiB, 13183593.7 MiB, 12874.6 GiB or 12 TiB per
second. That's more than your average hard disk, per second.
On a side note, does this sound plausible?
Since voxels has 'built in' lod levels, cant you use a much blockier/lower lod version of an object for physics and collision?
So instead of the current visible 64 voxels/mm3 they can use a non renderable 1 voxel/dm3 or something and use that as a rig?
Some simple springs or forces between those big voxels and you could have soft bodies and fracturing too.
That would require very little setup time, basically set some material parameters and that's it...it would 'skin' itself too since it's a hierarchy...
I think the biggest hurdle for voxel technology isn't simply showing that it can outdo current polygon rendering techniques, but straight-up outpace it.
Let's be real. Polygon rendering isn't going to stay static either. We've seen lots of hints that the next generation of consoles will be able to handle some pretty crazy stuff in terms of tesselation. Is the extra geometric detail from a voxel approach ever going to noticeably outpace the beefier and beefier output of polygon-crunching graphics cards, to the degree that anyone will ever risk re-training their entire production team? That's the real question, and the answer for it is pretty gloomy for voxels.
It just seems much more likely that polygon rendering capabilities will continue increasing at such a rate that the geometric differences between a polygon scene and a voxel scene will be hardly noticeable.
On a side note, does this sound plausible?
Since voxels has 'built in' lod levels, cant you use a much blockier/lower lod version of an object for physics and collision?
So instead of the current visible 64 voxels/mm3 they can use a non renderable 1 voxel/dm3 or something and use that as a rig?
Some simple springs or forces between those big voxels and you could have soft bodies and fracturing too.
That would require very little setup time, basically set some material parameters and that's it...it would 'skin' itself too since it's a hierarchy...
There are a couple of Patent *applications* (note with extreme prejudice that these are specifically 'applications' and not actual Patent grants) with the CEO's name on them, but no documentation is included, not even an abstract outlining the premise of the 'invention'. It's suspected therefore that what he's actually doing is developing a paper trail of "origination" for an idea he'll sell after building the hype around it. If this were real tech and as revolutionary as it's implied, he'd have fully fledged patents granted in his name from 2004 when the first application was made.
Ahh true, he could be building a case not to sell his "invention" but to use to file against patent infringement, basically he could be a patent troll in the making... or probably will turn to that once he figures out no one is biting. Watch out anyone with similar inventions. Beware the broad sweeping patent.
50 moon bucks says he sells his patent to Intellectual Ventures or someone like that if not hire himself a lawyer and start sewing anyone who uses the word edge... I mean voxel...
He also got what is being touted as the largest grant from the Australian Gov in Oct last year of nearly $2 million AU dollars, so he's not short on funds to register a Worldwide Patent on this tech. One can quite easily speculate the hype/marketing is simply to ensure the returns what is otherwise vapour-ware (his product, not voxel tech).
I think the biggest hurdle for voxel technology isn't simply showing that it can outdo current polygon rendering techniques, but straight-up outpace it.
Let's be real. Polygon rendering isn't going to stay static either. We've seen lots of hints that the next generation of consoles will be able to handle some pretty crazy stuff in terms of tesselation. Is the extra geometric detail from a voxel approach ever going to noticeably outpace the beefier and beefier output of polygon-crunching graphics cards, to the degree that anyone will ever risk re-training their entire production team? That's the real question, and the answer for it is pretty gloomy for voxels.
It just seems much more likely that polygon rendering capabilities will continue increasing at such a rate that the geometric differences between a polygon scene and a voxel scene will be hardly noticeable.
It's really hard for me to think of a major evolution in polygonal rendering tech beyond the tesselation in DX11. When you have the potential to generate geometry resolution, real-time, in such a controlled manner, what's left?
Seems logical that any new advances in 3D engines will focus on materials and shaders, lighting and shadows, or post-processing effects. Texture filtering, texture map compression, better LOD systems, occlusion, more integrated physics, etc. Still plenty of areas for game engine to evolve outside of merely rasterizing 3D form.
And of course tools. The tools have a long way to go to catch up with everything else.
More hardware-driven flexibility has always been the path, moving from fixed pipeline to a shader pipeline, with shaders gradually getting more and more flexible.
since i saw this video circulating all over the place lately, i figured itd be something incredibly mind blowing. it is. but i just dont see it happening in games for a long time. im happy enough with hardware tessellation right now.
oh yeah. and has anyone mentioned how much this guy sounds like a total dbag? :P
Replies
Release a video of his tech demo, talk a bit how 'polygon companies' are evil and banding together to make games look flat.
What ever happened to the sdk he was going to release last time? I'm not saying that a voxel based graphics could be the future. I just can't see this company being the ones to do it.
Its just the same marketing spiel they've always been giving. Just with newer art and no sign of release.
2003:
I'm betting that they promised that they where just *months* away from getting there engine to market back then too.
Basically this.
We're supposed to be some of the most open-minded folk in any industry, yet new technologies get this much flak?
Sure, they may not have all the answers yet. As a society, we know so little about the universe so that is a given.... which is why I always hit my head against a wall when people willingly CHOOSE to be close-minded to new ways of doing things.
Change is how we became what we are today. What would have happened if as a society, the development of the computer was just ignored and ostracised? They'd still be trying to market the technology now to a select group of people.
I just think it is funny that they come back, a year later and show MASSIVE improvements over what they first showed.... yet they still receive the EXACT same flak.
Why hasn't this technology taken the industry by storm?
No, it has little to do with physics, animation, any of that. That can be worked around because we are humans. We find solutions to EVERYTHING. We don't stop just because 'aww it didn't work', we PUSH ahead and that is how we are what we are today.
The issue is that people are simply not comfortable with change as a whole. Now with the polygon conversion tools, that is closer to becoming moot. As well as the fact that point clouds are just geometry; point-cloud rendering systems can be put into any rendering engine. Additionally, you can paint directly on the model or unwrap, and that texture data will also be transferred to the point-cloud models as seen in their demonstration.
They also showed animated point clouds in their last video. Seems as if everybody wants to just make an instant decision on this technology without actually doing any research on the subject.
[ame]
Don't care how bad it looks. That is point clouds being animated. So, major downside for point clouds #1 is no longer relevant, as it never existed as a concern to begin with.
Physics: Atomonage Engine has already displayed that.
Here is the true potential in regards to the graphical fidelity of point clouds.
[ame]
19 seconds onward. Sure, it may be raytraced, but it can just as easily be rasterized and extra shading power could be dedicated on the GPU since the geometry can be rendered entirely on the CPU... on one core at that.
Every game these days seems to use normal maps that reference extremely high-detail models. Digital sculpting is swiftly becoming the de-facto method for adding detail to game worlds. Almost everything is starting off as high-detail digital sculptures.
A tech demo like this one is just showing someone's efforts to bring full-on digital sculpting into the rendering space. I don't see any reason why this couldn't be used as a foundation for modern game rendering. Sure, there are features that need to be addressed. But the same was true of how 3D game rendering was originally handled. This is simply the first step in a different direction.
I still remain skeptical. Too many unanswered questions. How are they dealing with materials and shaders, lighting, animation. Changes in environments through destruction or something as simple as a door. How does this work with modern 3D hardware being so geared towards rendering triangles.
Clearly, something like this will one day happen. Maybe this is it, maybe not. It will take more than whitepapers and recorded visualizations to prove to me that this is more than a money trap for VC.
But more importantly, we still have a long way to go with tools to fully utilize even this generation's ability to render a comparibly limited amount of polygons. Projects are still struggling with constructing an environment on schedule and in budget within the current technology. Throw 'unlimited detail' in the mix, and you're talking about much higher than film level of quality environments. Beyond what's needed for the shot; a fully interactive 3D environment.
Sure, there's some tradeoff with not having to jump through the extra hoops of projecting high-poly models on to low-poly assets...but that doesn't make up for having to model invidual pieces of sand and blades of grass to populate your worlds. A lot of these tricks for dealing with limited polygons in a game engine are also incredible time-savers.
Unwrapping for textures, bones for animating, collisions, physics... we had to come up with all of that ourselves to make 3D viable for games.
The same thing simply needs to be done here. We've shown that as humans, we can pretty much solve any problem.
Also, remember that the scenes shown are being rendered on a single-core CPU... completely in software rendering. Think about how much processing power that leaves developers with today. Even on a console such as the Wii.
Here you go:
[ame]
Their newer demonstration in the first post is certainly much better.
This diatribe is so full of assumptions that one would suspect you have a personal, vested interest in this company and their technology being taken seriously and at face value.
Big difference between skepticism and close-mindedness.
What? Why? Last i checked it wasn't called free detail, you still need artists.
There's this growing consensus that you must achieve the same level of detail as what was seen in the unlimited detail tech demo videos.....for goodness sakes - its just a tech demo video...not a "YOU CAN'T MAKE GAMES WITH HAVE LESS DETAIL THAN THIS" video.
The less whimpering and more doing happens the smoother and faster the transition will occur and a resultant win-win for all.
Or that maybe I am open to change, or have an open mind, or as an artist, I can appreciate the changes it would bring?
There is a very small difference between skepticism and having a closed mind. They both lead one to stick to their ways and do not allow one to grow, change, and adapt.
I say that because a true skeptic will search out the information... but most people who say they are 'skeptical' are saying that because they are closing their mind to the idea, and refuse to learn anything more on it.
Trying to attack the individual because you refuse to acknowledge the actual information.... classy. :thumbup:
Euclidion will really need to push hard to get someone to make a cool game and get this working on existing hardware for anyone to take a serious look.
Also, point B which everyone is already talking about is the presentation. He's making big claims using a style I generally associate with liars and mentally unstable people. I think pretty much everyone has picked up on that to some degree. Makes this hard to believe in, but I do believe they (him?) do have an actual product.
But, his presentation style is certainly effective in getting the name/idea out there and talked about.
Also the guy talking pisses me off.
And unlimited detail doesn't always make a game look good.
Memory is the other issue with the animation setup, since you're storing each frame of the animation, you have to store the mesh for each frame. Being a point cloud, that will eat up a decent amount of memory per frame. Now multiply that by say even just 30 animations and you have quite a large file. Toss in the fact you'll need more than 1 character on the screen and forget about dynamically loading an animated point cloud into memory on the fly isn't really feasible due to the amount of time required to load from the hard drive (getting better with ssd's, but most people don't have them still).
What information? There simply is none. There are no live demos, no patents, no sdks, no technical reports, nothing to even imply that the product actually works. So again, what information are you refering to?
What if you use traditional skinned and animated polygon geometry that serves as a cage driving point-cloud data deformation?
This is not some new alien tech, others have done similar things, people working in the industry has touched on this, even Carmack who is one of the visionaries you speak of, even he is man enough to admit the limitations of mega-texture, where it shines and where it doesn't.
Point cloud data is not geometry, you cannot simply put it into a rendering engine and get geometry from it.
I would agree with this, not in the way you think though.
As someone mentioned, animated, not skinned.
One of the downsides with this is that the geometry for each object has to be stored somewhere, the unlimited detail isn't generated from thin air, these are massive objects. mesh-per frame animation would just be silly in this scenario.
It just works magic by instancing the same meshes.
Atomontage is technically a very different beast from unlimited detail.
No, that's exactly the same use we've seen of point cloud data before: a huge amount of static point cloud data and a smart way to read it, you don't simply just share work in the computer, you have to take into consideration how it actually does things.
Game engines are a series of compromises. You could put a Zbrush model into UDK and talk about how amazing it looks, but it's going to consume your budget. You could use 2048x2048 maps on everything, but it's going to impact your memory budget.
These guys are shortsighted and arrogant. Don't compare your rendering engine to fully featured games, and then claim how ugly and stupid they are.
And about the animations, skeletal deformation is still going to be preferable to point cloud animation. Animated point cloud data is pretty much what Quake used and is the same as per vertex animation? But instead of drawing faces between the points and using less points you use so many points that its so dense you can't tell there aren't any faces between the points?
There is a point when it becomes so dense and so much data is floating around you're better off using less points and drawing faces between them.
It seems kind of weird that anyone would think that a half finished system that strips out the faces, packs the dead space with points and provides less tools and less functionality would get any kind of positive feedback.
Artists are visionaries yes and we 3D monkeys embrace new and weird stuff all the time but part of being a smart monkey is being able to discern what is useful from what is pointless and counter productive.
************************
Well, from what they show, it looks very impressive.
However, everything is static and appears to be colored by the little "atoms" themsevles. Looks like there aren't any textures involved, I could be mistaken though.
It certainly is an interesting new technology and I'm eager to see how this turns out.
However, I noticed how they/he talks about the rest of the industry and years of improvement like total garbage. I love the part where they show footage from Crysis 2 and tell us that their technology could make it look thousands times as good
Anyone noticed how he points at backgroundobjects and tells us that these are just fake cardboard meshes. What does he expect artists to do? Spending weeks on a backgroundobject that will be seen like 5 seconds in the game?
Artists get payed and they are not cheap, studios have releasedates and publishers no unlimited budget to invest unlimited time in a game with unlimited details.
How is lighting supposed to work with this? Shadereffects? Any possibilitie to texture these models? Animation, dynamic objects? (Notice how absolutely everything is static in their preview scenes)
Well, I'm no professional and might not be up to the latest knowledge and tech'.
But this looks a bit hard to believe to me. Not the technology itself makes me doubt it, but the presentation.
Notch's tweets can move entire games, so imagine the PR-damage this blog will cause :P
Can't see that as being efficient. This particular tech is more akin to a search algorithm...which doesn't lend itself to dynamic data. This is data that is being managed and organized in very particular ways beforehand to make the method of finding the right points to display as quick as possible.
Certainly, an engine could use some sort of hybrid approach. Building the static environment out of point-cloud and any dynamic geometry could use polygons. But here's the thing...at some point, throwing more triangles at an object really doesn't do any good. When the triangles are smaller than pixels, there's no point in rendering more triangles. If they're off-screen, there's no point in rendering them. If they're occluded, there's no point in rendering them. The end goal of a visual media is not to recreate reality, it's to create a convincing illusion of reality. Even Hollywood doesn't work with unlimited detail, and they've got one of the best rendering engines every created.
Right now I believe that by the this type of tech is truly ready for production the traditional polygon rendering engines will either have the raw hardware horsepower or developed enough new tricks to be able to render as much detail as you could actually need. I could very well be wrong about that, but with almost the entire the industry devoted to improving the current method over throwing everything out and reinventing the wheel, I'd be very surprised if polygon rendering was replaced anytime soon.
Either way, it really doesn't matter to the artist. What bottlenecks art in games now is generally not an inability to throw more detail on the screen. It's how long it takes to create those high-poly assets and the hoops one has to jump through to get those assets into the current engines. I feel we've already hit a bit of a ceiling on this...that modern engines are more than capable of rendering detail that most companies simply don't have the time, manpower, or technical proficiency to fully utilize. So it essentially boils down to the tools for creating these assets and the pipeline for getting an asset into a game.
People can get all tingly about how exactly about how unlimited visual detail is rendered, but ultimately it's how all that unlimited detail gets created that needs to change first.
Obviously, he's close-minded.
And noted the projects that should be getting the attention, projects that don't pretend do be something they're not.
Another one will be along shortly.
And about Notch's blog post, Ah-HA!
They say they use 64 'molecules' per mm3, That would mean
they use 64,000,000 for a single square meter. <br><br>One 'molecule' at least
uses integer-datatypes as positions for x y and z, which would be 4 or 8
bytes each, then they would need something to discribe the color,
lightning and the reflection of the molecule. I guess about 50 bytes per
molecule.<br><br>50x64 000 000 = 3,200,000,000 bytes of data per frame per
m3.<br><br>Say you have a room of 3x4x6 meters, and you want
to render it at 60 fps. Let's do the math...
3200000000x3x4x6x60 = 13,824,000,000,000 bytes per second to send
through your computer.
<br><br>That equals 13500000000 KiB, 13183593.7 MiB, 12874.6 GiB or 12 TiB per
second. That's more than your average hard disk, per second.
Computers are fast, but not THAT fast.
On a side note, does this sound plausible?
Since voxels has 'built in' lod levels, cant you use a much blockier/lower lod version of an object for physics and collision?
So instead of the current visible 64 voxels/mm3 they can use a non renderable 1 voxel/dm3 or something and use that as a rig?
Some simple springs or forces between those big voxels and you could have soft bodies and fracturing too.
That would require very little setup time, basically set some material parameters and that's it...it would 'skin' itself too since it's a hierarchy...
Let's be real. Polygon rendering isn't going to stay static either. We've seen lots of hints that the next generation of consoles will be able to handle some pretty crazy stuff in terms of tesselation. Is the extra geometric detail from a voxel approach ever going to noticeably outpace the beefier and beefier output of polygon-crunching graphics cards, to the degree that anyone will ever risk re-training their entire production team? That's the real question, and the answer for it is pretty gloomy for voxels.
It just seems much more likely that polygon rendering capabilities will continue increasing at such a rate that the geometric differences between a polygon scene and a voxel scene will be hardly noticeable.
Try telling me this realtime tesselation demo isn't a wildy more compelling demo than all the voxel stuff ever released:
[ame]
It's possible, it just isn't very fast at all.
http://www.npr.org/blogs/money/2011/07/26/138576167/when-patents-attack
50 moon bucks says he sells his patent to Intellectual Ventures or someone like that if not hire himself a lawyer and start sewing anyone who uses the word edge... I mean voxel...
It's really hard for me to think of a major evolution in polygonal rendering tech beyond the tesselation in DX11. When you have the potential to generate geometry resolution, real-time, in such a controlled manner, what's left?
Seems logical that any new advances in 3D engines will focus on materials and shaders, lighting and shadows, or post-processing effects. Texture filtering, texture map compression, better LOD systems, occlusion, more integrated physics, etc. Still plenty of areas for game engine to evolve outside of merely rasterizing 3D form.
And of course tools. The tools have a long way to go to catch up with everything else.
oh yeah. and has anyone mentioned how much this guy sounds like a total dbag? :P
Nope. Just you so far.