I'd be much more inclined to believe this was the future of game rendering, if that fella were moving around, breathing, and showing the kinds of deformations we take for granted in polygonal rendering.
@olilittle Trace rays in the scene onto an animated simple bounding mesh, which is then used to translate the rays into mesh coordinates in the base pose, that are then traced against the original static voxels. So the voxels don't actually move, but your view of them does. Need to take care to deal with intersecting animated mesh - trace twice through the voxels.
@Azyashi It might be worth mentioning that the upcoming id Software's Rage has many of the same non-moving restrictions (as do a very many other games). In most games, only certain objects are movable - and I provide a plan for that in my siggraph presentation.
Voxels are nothing new. They were being used on a few games on the SEGA Saturn. The thing with vocals is that animation algorithms out there right now simply aren't fast enough for good animation. That's why they're mainly used for terrain.
Also... is there a kind of "sweetspot"? for animations. The "race to the top" thing is quite boring to be honest. See how minecraft with it's super big voxels has become popular. Photorealism is a boring goal.
What if you used animated characters, say 128 pixels tall, but with fast animation rates? And each pixel was lovingly created with the same attention to detail they had in the good old SegaGenesis/SNES days?
@noobenstein - Minecraft does not use voxels to render the terrain. It's uses a pretty standard polygonal renderer. It just happens to use a blocky 3D grid to _define_ the terrain, which is a look people normally associate with (low-res) voxels.
@RFC3514 Notch uses voxels to define the shapes of the blocks. If he didn't use voxels, it would be impossible to render all those cubes you're able to see on-screen.
@Haroids Why would it be impossible? The poly-count isn't very high at all compared to current generation games. And even if that is what you mean it wouldn't be impossible, just slow to render. Unless you mean impossible in some other strange way. Care to elaborate?
@Tableuraz This was done in 2008 (3 years earlier), with a GPU which was 8 times less powerful than the one UE3 used in their latest tech demo. ;) And the model IS textured with a light probe from Paul Debevec.
This is so much different dude. This is huge! This will change everything to do with gaming. This will produce photorealistic games all while making it very easy to run.
@Tableuraz individual voxels can be individually colored and shaded to act as a texture. Rendering untextured voxels is essentially the same as rendering them colored. On top of that, this is an extremely detailed, production quality model, not something that would be used in-game. If this were based on polygons, and modeled at the same density, UE3 would refuse to render it at all.
UE3 uses a low-poly cage with an object normals map. Game-ready normals maps are NOT this resolution nor do they dynamically scare. It's a completely different technology.
I think voxels are really cool and we're certainly doing things with the technology that show its potential for replacing polygon rasterizing. The problem that remains however, is that trying to animate a voxel data set is like trying to animate a Lego man made out of thousands or millions of Legos. In the next few years, we're going to have to find a way to resolve that. I think that some sort of polygon-to-vector conversion thing might do it, though we may find more practical methods.
i'm not as impressed with this as i'd like to be. yes, the idea of never ending levels of detail are great, yes it's running at 60 fps- however, this is just one object, with one light source, and that object is static. show me the fps when it's moving, attached to ai, in a room full of other animated, scripted critters with particles & light sources flying-tech 4's bane). not to mention that artists'll balk at having to learn a new pipeline. polys and megatexture seem to do the same thing.
This is cool and all, but I am waiting for deformable voxels, seems the technology is only fit for static rigid geometry. I see this being adopted in Environment art, but not character art.
The argument of voxels for destructable geometry in games is poor. In theory it might be easier to chop up voxels than polygons, but everyone ignores the huge memory requirements needed to store all of this dynamically generated data, as well as the difficulty of collision with dynamically generated shapes.
please, do you know any device to do voxel modeling using our own fingers?
I've done one called vitalino (search voxel vitalino on youtube) in an intial development state using two webcams and I would like to see similar technologies.
But in a game, why would you ever need to get that close to something that you're trying to kill??? You're not wanting to see his ear, you're wanting to shoot that fucker in half.
So all this talk of Voxels and infinite blah blah, it's all great and shit, but realistically what's wrong with polys? They work fine.
If everyone thought like you we'd still be using sprites... "FUCK IT WHY BOTHER WITH THIS NEW SHITTY POLYGON SHIT, SPRITES WORK PERFECTLY FINE! ITS THE FUCKING GAMEPLAY THAT MATTERS FUCK OFF!" Sounds good for the progress of games technology...
@Kewickviper 3D games are fine too. People can enjoy more realistic graphics while sprite based games such as castle crashers can also be enjoyed. Anyway this is done using Zbrush, the detail will never be this good in realtime 3d games this generation, unless transformed into a displacement map.
Does it really matter that much? I mean, it IS a gorgeous model... Voxel, polygon... If it looks good, we're all buying it... And, remember that technology is not a linear thing, we can evolve by different ways.
There was some old Delta Force game that used voxels - it didn't look so hot at the time but it let you have lonnnnnnnnnnng draw distances. So this kind of technology isn't just for pretty detail when you get in close, it can give you good detail when you're very far away too!
@staphinfection what's wrong with polys? for one ,as the resolution of the image goes up, polygons become a sub-optimal solution for storning the data, actually causing MORE work than it saves. A real time voxel engine could render any shape, at all, without having to convert it to a series of polygons first, and each pixel can be individually rendered (perfect for parallel processing) As for gameplay too, voxels enable cool little things like fully destructable environments, etc..
True hardware supported displacement mapping will superceed this, It has the benefits of the fine detail and animation capabilities and is more compatible with the current pipeline for 3D asset creation.
if nvidia gets thier 'supercomputer architecture' of the ground and working, more complex tasks could be achieved with paralell processing, and the billions of ants principle would make effects like this far faster than current hardware - but there will allways be a limit to what we can do we need a new way of codeing
For dynamic models, you can probably pull it off easily with an iterative feedback mechanism with the screen space. Meaning, transform visible voxels from previous scene, see if LOD is satisfied, if not, grab some more voxels from the octree, transform them, and continue until LOD is satisfied.
It will take another 3 to 4 years before Voxel Rendering hits retail. The GPU power will reach the point where you can render voxel at 60fps. But in the year 2010, it's still not possible.
I see a lot of people argue that voxels will not be the future of gaming due to hardware limitations.
But think about how fast our technology is evolving, and has evolved. This just ten years ago, when games like Turok looked "so real", now look at it compared to something of today. And think ten years before that.. Atari..
We'll get there in time. If not on the next generation of consoles, then the one after that, along with VR goggles :-D
Normal mapping is a geometry approximation which fails with a moving camera, at oblique angles, representing surface concavities, and incorrectly shows sides of surfaces you cannot actually see causing lighting artifacts. Normal mapping is also typically associated with polygons, and polygons have limitations on how many you can push through the hardware.
SVOs on the other hand are only limited by how much hard drive space you have and the compression algorithms you use to reduce the amount of hard-drive space you use. You can represent a nearly unlimited amount of geometry and geometric complexity.
I should mention that SVO is an approximation as well. While it doesn't fall prey to many of the typical approximations, if falls prey to another. For one, voxels enumerate a surface at discrete points. For example, a standard definition TV signal is blurry where as a HD signal is sharper and more accurately represents the original scene. As such the resolution of the voxels affects how accurately it represents the original scene.
A good question along the same vein which really gets to the heart is, what can I do with SVO that I can't do with polygons?
I think that SVO is a simplifying factor. It solves LOD, streaming texturing, streaming geometry, uv unwrapping, etc... all in *one* single algorithm. A giant hammer of which to make game development easier and higher quality. The end result of voxels is what many other algorithms struggle for and what voxels achieves relatively effortlessly.
I agree, in part. But, there are always going to be issues. I see voxels as an answer to liquid simulations and volumetric effects. As well as surface practuring. A more genuine issues is still the process of creating structured characters. which would require a 'surfacing' effect to the outer voxel points, defining a shell, or even the idea of positive/negative spaces for physics simulations, which might be heavily reliant on a system similar to BSP...
@zelexi They couldnt be dinamically rendered so far surfaces use less voxels than near ones? For example, using the same amount of pixels on screen in voxels. So when you get near, the models always looks "good" and never pixelated, maintaining always the same amount of voxels on screen. Could be that possible?
@zelexi this also true with polygons though. the difference being that mesh detail, texture resolution, texture filtering techiniques and more is also affecting the end result not just output resolution and input object resolution.
@zelexi That definition of a normal map is a little false. Normal maps are agnostic of geometry and instead simulates the reaction between surface variation, light source and observer. Normal maps are a an efficient optical illusion, which is true of all good pixel shaders.
@Gusdor I actually entirely disagree with that. Normal mapping defines the normals across a surface. It very much is a geometry approximation. How you use those normals in a shader are irrespective of what the data represents.
@zelexi I'm no graphics programmer, but I think Gusdor is technically correct - certainly a height/displacement map would be a geometrical approximation - a normal map is derived from this, but looses certain information, i.e. a normal vector parallel to the surface normal would be assumed to reside within the polygonal plane, whereas in real life it may be displaced above or below the plane. The normal map in this case gives no clue as to the real geometry it was derived from.
@antzrhere Using a height/displacement map to make a normal map is using an approximation of geometry to make another approximation of geometry. Two levels of approximation are still an approximation. The correct way to make a normal map (with fewest levels of approximation, and maximum quality) is to make a very very high poly mesh, and a low poly one. Then project every pixel on the texture from the low poly mesh onto the high poly one.
@antzrhere I should also mention that a height/displacement map of the same resolution as a normal map, cannot actually represent all values of a normal map. It can only approximate a normal map. While this is normally good enough of an approximation these days, back in the day when normal maps were first used with very low poly meshes, this was a very horrible approximation.
@zelexi I agree,they're both approximations, and I can imagine displacement maps suffer more as the degree of displacement increases with constant resolution.How I see it though,displacement maps could be *crudely* converted back into a polygon mesh (of course, this would only hold true if there were a single topological layer -grooves within grooves cannot be represented with a single displacment map),but a normal map couldn't on it's own.
@zelexi No, your correct.It is a geometric approximation -I suppose geometry as a definition encompasses volume,points,lines AND angles,so I must confess,I was originally wrong.Furthermore an 'approximation' has been defined as:"an inexact number, relationship, or theory that is sufficiently accurate for a specific purpose" - so if it gives a sense of shape (be it believable or not) it is an approximation.In the context of re-constructing volume it is not an approximation,but thats irrelevant...
@zelexi Just a quick question -with 100% screen coverage(but not completely zoomed in),what sort of performance do you get on GPU (million rays/second)? About a year ago I tried to implement something similar -it worked nicely - ray-guided streaming,compression etc. but as it was in software,on a 3.9Ghz i7 I could only get 16-20Mrays/sec with 100% coverage - max 30M completely zoomed in. I know Nvidia SVO managed 55-60M and peaked at 110mps inc.filtering.Seems CPUs just cannot compete.
35,184,372,088,832 in total voxel space. The model is scaled to take up almost the entire space. There is still internal cavities that is collapsed and some outside space, however the total number is still staggering.
Note that I said "There is still internal cavities that is collapsed and some outside space" This probably reduces it by a few orders of magnitude, as well as there is some simple compression going on. I would recommend reading the paper. That large number that I stated is "total voxel *space*" not total number of voxels used by the model. That is a much harder number to determine without specifically implementing code to measure.
you should know yes :) and if you don't know, why should I read your paper... : )
I recommend you do a count of the voxels in the model, seems a pertinent bit of information in evaluating voxel rendering.
it seems that a 2gig model of a single character in datastructure that's impossible to animate, means that voxel rendering is not of interest and never will be. would you agree?
It is not impossible to animate. Just there there are better things to do with your time for x720/ps4. You may see animated voxels as I have seen research in this area. The method that I propose is to render the static objects with voxels (namely the world) and the dynamic objects still as triangles. Which makes sense as if you look at games today the characters look immaculate but the world can use some serious improvement. This gap has been closing and voxels is one possible end to that.
The 2gig level doesn't do FFT or Entropy encoding, so there is about an order of magnitude possible improvement. 2gig down to 200mb. For example, idTech5 when uncompressed textures used up on the order of 200gb for a 1 level (of which they have many). This compresses down to fit onto a few dvds. The same is likely to happen here, except that here we will have to fit onto a few blue-rays. So this technology is very feasible and practical for next-generation hardware I assure you.
this is a tech demo. The real thing is still a few years away, by then Intel's Larabee will be out and maybe the entire graphics industry will recieve another overhaul similar to what unified stream processors did to the old vertex/shader pipelines.
Not really, 60fps for this is fking amazing. Have you ever tried coding voxels. Performance is the craps on them. This is impressive, and your forgetting the common scale, when a game runs at 60fps *only, pfft, that dosnt mean 2 objects will be 30fps, often a super simple game example with a complex engine can render a cube at 800fps. And u think omg only 800fps for nothing onthe screen thats bad. Yet you can add 10,000 objects and it still will maintain 40fps or so.
Absolutely they show stuff like this without application in their mind, because they know that a lot of people think like you, and think thing like "such a great man! he has invented the future of video-game!".
But in fact: no.
Voxel is unusable (through currentl real-time technologies) for advanced animations and interactions.
Eventually usable for really static environnement (allow very detailed things with easy advanced lightning). Currently noway for anythng else.
How many light sources? Seems brightest source is upper-left, yet shadow is lower-left. If there was also a brighter upper-right source, this could make sense... Otherwise, I'm scratching my head in confusion.
Could be two. The one to the right is at a lower angle so it doesn't shade the side of the body on the right hand side.
In all honesty I don't know what the point is in this Voxel technology unless it actually offers possible improvements. With polygons catching up in terms of detail, what's the point in researching Voxels?
Project Offset looks very much like this, by the way.
Would be cool if a Computer had Voxel and Polygon-specific rendering at the same time, though. Voxels can, potentially, change gaming future as we expect it?
zoomed in and rotating is technically a worse situation, because the more zoomed in you are, more of tree is not loaded in when you rotate around seeing pieces of the model you have never seen causing an even worse immediate loading situation.
Cryengine 2 uses voxels for some geometry. As the terrain is heightmapped they use voxels to make stuff like cliffs, caves and rocks in their terrain as the heightmapped surface cant.
Voxel is a kind of 3D models which can be directly display like voxel, in fact it's even the interest!! because this kind of display have advantages, like very high performance for very detailed models (thanks to a built-in adaptative level of detail, automaticly combining both 3D model and texture-resolution adaptation in the same time) , and easily lightning (with dynamic shadow, etc).
Prob: currently no way for advanced animation and interaction. So static environnement only.
comments on this thread are silly; voxels could perfectly be used for all 3d entities (terrains, characters, etc), the problem is current hardware is designed with polygons in mind; technology created with voxel graphics acceleration exists, namely in the medical field for TAC scans visualization and whatnot, we just need a change of paradigm, maybe in the form of a gaming console focused on voxels... ...
True, there are some places that would make it for voxels a problem in games. However that's why I think the future is gonna be more of a hybrid, Polygons still will hang around as well as mainly software render/programing/shaders taking over that will be accelerated weather it be voxels or whatnot.
And having a good voxel modeling program would just be something like Zbrush but using voxels. And polygon -> voxel is always a option too, or in some cases maybe voxel->poly for destruction's.
@Nouseeker problem with voxels is although they make beautiful rigid still images. hardware just cant cope with that level of detail when you introduced movement.
@Nouseeker But how can you calculate the matrix deformations on millions of voxels at an acceptable realtime performance?
The voxel technique seems great for static meshes. But for anything that needs to deform, I think traditional Polygonal systems plus hardware tessellation are the better choice for the foreseeable future. (until we have hardware that can deform massive voxel datasets.)
@Nouseeker Voxels can't be animated. At least not animated well. Show me one example of a voxel character that is not animated with frame-by-frame. Basically, voxels mostly suck for several reasons -- A) No way to animate a character well -- B) Filesize grows exponentially larger as you increase detail -- C) Difficult to texture (requires paint directly on model, no uv / unwrap)
Not going to happen, all the technical stuff aside, we've just perfected the polygonal techniques way too far to consider a switch, the amount of time and effort required to change factors for the simple benefit of seeing more detail on extreme close-ups is just not worth considering that in the end it is the overall experience of the game that counts, not what the character looks like if we put their fingerprints up to a 200x zoom.
@Shitsocash Thats a valid criticism for characters. However, it this was world geometry and not a character thats entirely different story. And really, the tech is for world geometry, and not really characters all that much. I only used characters for the siggraph demo cause thats what I had.
@Shitsocash This is the same logic that is preventing us from colonizing and making material use of other planetary bodies. Because we choose to take the least short term cost for the most conservative reward regardless of how much it will cost us in the long run. Hopefully Asteroid Toutatis will cure our lack of foresight!
@Nouseeker It's not just a rendering problem. There's also the issue of animation. While voxel solids can be animated perfectly well, the tech doesn't lend itself well to the traditionally mesh-based deformations needed for soft bodied, flesh, cloth, etc. Until this is solved, I think we're pretty much just going to see hybrid engines.
I believe animation is still possible, I remember lattice skinning being mentioned somewhere. Atomontage has an example of a hybrid polygon voxel engine.
This demo uses a volumetric octree, or a representation of space in which each node of the tree is divided up into eight parts, and as you descend down the tree you get a more refined approximation of the original, high-resolution source material.
If you tried doing this brute-force, using the original polygonal mesh, it wouldn't be real time. It would be a slideshow.
I see, but that's not what I meant by that. RoboDonut said that this voxel model was divided in 8 parts and the more you got closer to it, the more defined it became. In actual games and movies, the closer you get to a texture the more you see the pixels and it eventually becomes very blurry. Can a similar LoD approach be done with textures? That would be very cool if one day it'd be the case.
yup :) this is kind of built into the technique, as there really aren't textures anymore. Each voxel has both position data and texture data inside it. Similarly, there is no need for normal maps for geometry approximation anymore as we can represent this geometry directly.
That has already been done (Paged virtual texturing, as seen in Quake Wars and rage), but it would actually be moving backwards. With THIS technology there would be no more need for textures at all - all detail would be (or at least COULD be) modeled
Im not so sure that textures are done with, you obviously need some kind of colour(texture) information or all youll get is grey or single toned models, but thats why I think procedural generation for both texture information and these voxels would be the future, assuming voxels are able to be animated, if not then that obviously limits Voxel use to static geometry, which is becoming less and less used in todays games
They are only going to use this tech for the enviroment. Characters will still be made out of polys. Unless the next gen hardware gets better then everyone in the gaming industry thinks it will.
They are only going to use this tech for the enviroment. Characters will still be made out of polys. Unless the next gen hardware gets better then everyone in the gaming industry thinks it will.
I'd be much more inclined to believe this was the future of game rendering, if that fella were moving around, breathing, and showing the kinds of deformations we take for granted in polygonal rendering.
GotNextVideo 2 months ago
99999999999999999999999 voxels in this model... maybe more
gabrielchcosta 2 months ago
i wouldnt want to go "all night" with that beast.
i think the music is unfitting.
ImTheWackhead 2 months ago
@ImTheWackhead So true. I need to fix that.
zelexi 2 months ago
Tell me, what do you plan to do about animating said voxels?...
olilittle 3 months ago
@olilittle Trace rays in the scene onto an animated simple bounding mesh, which is then used to translate the rays into mesh coordinates in the base pose, that are then traced against the original static voxels. So the voxels don't actually move, but your view of them does. Need to take care to deal with intersecting animated mesh - trace twice through the voxels.
ZanyRocket 2 months ago
It's like a photo, but then 3D :o
MrEggBass 4 months ago
I wonder how it would look after being in the hands of a great texture artist.
jensi18 5 months ago
Unlimited detail atomic graphing is the future and is close
beutifulchaos3 5 months ago
@beutifulchaos3 Sure! As long as you don't mind having no movement or collision detection. Both of which are very difficult to do with Voxels
Azyashi 5 months ago
@Azyashi It might be worth mentioning that the upcoming id Software's Rage has many of the same non-moving restrictions (as do a very many other games). In most games, only certain objects are movable - and I provide a plan for that in my siggraph presentation.
zelexi 5 months ago
Comment removed
infraball 5 months ago
@beutifulchaos3 No.
RagingMonocle 5 months ago
whats the song name?
ty for anwering
Ceranoa 5 months ago
@Ceranoa
Circ - All Night
Nima5124 5 months ago
Doom 5
KaiSoDaM 5 months ago
Voxels are nothing new. They were being used on a few games on the SEGA Saturn. The thing with vocals is that animation algorithms out there right now simply aren't fast enough for good animation. That's why they're mainly used for terrain.
HedgehogStudios1 5 months ago
Can you get animations going?
Also... is there a kind of "sweetspot"? for animations. The "race to the top" thing is quite boring to be honest. See how minecraft with it's super big voxels has become popular. Photorealism is a boring goal.
What if you used animated characters, say 128 pixels tall, but with fast animation rates? And each pixel was lovingly created with the same attention to detail they had in the good old SegaGenesis/SNES days?
noobenstein 6 months ago
@noobenstein - Minecraft does not use voxels to render the terrain. It's uses a pretty standard polygonal renderer. It just happens to use a blocky 3D grid to _define_ the terrain, which is a look people normally associate with (low-res) voxels.
RFC3514 4 months ago
@RFC3514 Notch uses voxels to define the shapes of the blocks. If he didn't use voxels, it would be impossible to render all those cubes you're able to see on-screen.
Haroids 4 months ago
@Haroids Why would it be impossible? The poly-count isn't very high at all compared to current generation games. And even if that is what you mean it wouldn't be impossible, just slow to render. Unless you mean impossible in some other strange way. Care to elaborate?
buffynajtor 3 months ago
I can't access the paper using the link from the description, it seems to be offline.
Nocturnal321 6 months ago
@Nocturnal321 Yeah, it does look down. I'll see if I can find a mirror. (Please anybody else reply if you already know of one)
zelexi 6 months ago
@Nocturnal321 I put a mirror up in the description
zelexi 6 months ago
This is kinda like a tech demo for IdTech6 and Doom 5 :D
gevelegian 6 months ago
terranova for pc back in 96 was using voxels. sorry to burst your bubble but this wont change anything.
LOLOLWNED 9 months ago
search for animated sparse voxels, radulphi has some interesting videos on the topic.
1776Phoenix1986 10 months ago
Wow, 60 FPS for an untextured model!!! Seriously, the UE3 makes quite the same quality flawlessly...
Tableuraz 10 months ago
@Tableuraz This was done in 2008 (3 years earlier), with a GPU which was 8 times less powerful than the one UE3 used in their latest tech demo. ;) And the model IS textured with a light probe from Paul Debevec.
zelexi 10 months ago 9
@Tableuraz Also, UE3 wasn't 60 fps on consumer grade hardware.
zelexi 10 months ago 6
@Tableuraz The point is that it is made of voxels, not polygons. This is rather good for voxel technology.
cowtrix 9 months ago
@Tableuraz You're tragically uninformed.
bonchbonch 9 months ago
@Tableuraz
This is so much different dude. This is huge! This will change everything to do with gaming. This will produce photorealistic games all while making it very easy to run.
streetridaz 9 months ago
@Tableuraz individual voxels can be individually colored and shaded to act as a texture. Rendering untextured voxels is essentially the same as rendering them colored. On top of that, this is an extremely detailed, production quality model, not something that would be used in-game. If this were based on polygons, and modeled at the same density, UE3 would refuse to render it at all.
WerdTalker 9 months ago
@Tableuraz Someone doesn't know what normal mapping is
abrythenabry 8 months ago
@Tableuraz
It's not the same quality....
UE3 uses a low-poly cage with an object normals map. Game-ready normals maps are NOT this resolution nor do they dynamically scare. It's a completely different technology.
nhurm06 8 months ago
смысл выполнять весь объем пикселями?! будущее за фракталами!
mechnotech 11 months ago
Something similar:
/user/AtomontageEngine
agenri1 1 year ago
this is what does 3d-coat software, you can download a demo
raterito 1 year ago
I think voxels are really cool and we're certainly doing things with the technology that show its potential for replacing polygon rasterizing. The problem that remains however, is that trying to animate a voxel data set is like trying to animate a Lego man made out of thousands or millions of Legos. In the next few years, we're going to have to find a way to resolve that. I think that some sort of polygon-to-vector conversion thing might do it, though we may find more practical methods.
CodyRicheson 1 year ago
It's been over a year, I wonder how it looks now..
jhtrico1850 1 year ago 4
Is this demo downloadable? I'd love to see one of my character models rendered with this SVO technique.
SSCrow 1 year ago
I think that, the first company wich actully give hardware support to this tech will epicly win over the others for good.
Bariudol 1 year ago
i'm not as impressed with this as i'd like to be. yes, the idea of never ending levels of detail are great, yes it's running at 60 fps- however, this is just one object, with one light source, and that object is static. show me the fps when it's moving, attached to ai, in a room full of other animated, scripted critters with particles & light sources flying-tech 4's bane). not to mention that artists'll balk at having to learn a new pipeline. polys and megatexture seem to do the same thing.
littlehorn33 1 year ago
and shaders? can apply to svo the bump, oclussion displacement mapping, global illumination withouth crash of fps of the raytracing?
crolus 1 year ago
Yes, it has nice details, but can it rotate and animate yet?
VCat2006 1 year ago
This is cool and all, but I am waiting for deformable voxels, seems the technology is only fit for static rigid geometry. I see this being adopted in Environment art, but not character art.
tddavis 1 year ago
Think about it this way...
You can run 37 MILLION VOXELS on a single Pentium 3 CPU, no overclocking, no nothing, at 60 fps.
A GTX 295 can't run a 37 million poly object even if you overclocked the hell out of it and paired it with ATI's newest cards.
huntpivot2 1 year ago
somehow this seems to be more "real" than polygon renderings. it seems to have a real body mass.
Adeltraut 1 year ago
The argument of voxels for destructable geometry in games is poor. In theory it might be easier to chop up voxels than polygons, but everyone ignores the huge memory requirements needed to store all of this dynamically generated data, as well as the difficulty of collision with dynamically generated shapes.
Slime01 1 year ago
please, do you know any device to do voxel modeling using our own fingers?
I've done one called vitalino (search voxel vitalino on youtube) in an intial development state using two webcams and I would like to see similar technologies.
Anybody can list some technologies for that?
thank you!
jabahpureza 1 year ago
insane!!!
FumingPoliticalPunk 1 year ago
But in a game, why would you ever need to get that close to something that you're trying to kill??? You're not wanting to see his ear, you're wanting to shoot that fucker in half.
So all this talk of Voxels and infinite blah blah, it's all great and shit, but realistically what's wrong with polys? They work fine.
IT S THE FUCKING GAMEPLAY THAT MAKES GAMES GOOD.
staphinfection 1 year ago
@staphinfection Artists need jobs.
Einhanderkiller 1 year ago
You're not trying to shoot the environment.
Gameplay is obviously important. But good graphics never hurt. They're something we should strive for as well.
Freecell82 1 year ago
wooosh
bubba22213 1 year ago
I'm not going to talk about such a big issue in 500 chars or less, because this tech definitely has its place.
However, I will agree gameplay > graphics in gaming. I still think pokemon for the gameboy and Super Mario Bros for SNES are the best games ever :>
chbrules 1 year ago
If everyone thought like you we'd still be using sprites... "FUCK IT WHY BOTHER WITH THIS NEW SHITTY POLYGON SHIT, SPRITES WORK PERFECTLY FINE! ITS THE FUCKING GAMEPLAY THAT MATTERS FUCK OFF!" Sounds good for the progress of games technology...
Kewickviper 1 year ago
I love sprite...
llollercoaster 1 year ago 2
@Kewickviper 3D games are fine too. People can enjoy more realistic graphics while sprite based games such as castle crashers can also be enjoyed. Anyway this is done using Zbrush, the detail will never be this good in realtime 3d games this generation, unless transformed into a displacement map.
majorhavoxg1 1 year ago
Does it really matter that much? I mean, it IS a gorgeous model... Voxel, polygon... If it looks good, we're all buying it... And, remember that technology is not a linear thing, we can evolve by different ways.
jackmcmorrow 1 year ago
How're the bookread and moviewatch treatin' ya?
MMMETALL 1 year ago
There was some old Delta Force game that used voxels - it didn't look so hot at the time but it let you have lonnnnnnnnnnng draw distances. So this kind of technology isn't just for pretty detail when you get in close, it can give you good detail when you're very far away too!
cactustactics 1 year ago
@staphinfection what's wrong with polys? for one ,as the resolution of the image goes up, polygons become a sub-optimal solution for storning the data, actually causing MORE work than it saves. A real time voxel engine could render any shape, at all, without having to convert it to a series of polygons first, and each pixel can be individually rendered (perfect for parallel processing) As for gameplay too, voxels enable cool little things like fully destructable environments, etc..
1Raptor85 1 year ago
@1Raptor85
Have you had a look at "unlimited detail" quite amazing!
DeathSlayer2 1 year ago
it's like google earth?
Corcoancaoc 1 year ago
True hardware supported displacement mapping will superceed this, It has the benefits of the fine detail and animation capabilities and is more compatible with the current pipeline for 3D asset creation.
ddnguyen278 1 year ago
if nvidia gets thier 'supercomputer architecture' of the ground and working, more complex tasks could be achieved with paralell processing, and the billions of ants principle would make effects like this far faster than current hardware - but there will allways be a limit to what we can do we need a new way of codeing
nazgul60 1 year ago
If anyone's still interested, the song is "All Night" by Circ.
philnolan3d 1 year ago
It's cool and everything. But...
1) there's no animation. Surely you're still way better off with polygons when it comes to efficient skinning and posing.
2) Animated or not, there's just no clear advantage to using voxels for this kind of thing IMO. Crumbling buildings or gouged landscapes maybe...
legogenius 2 years ago
Seems genius!
For dynamic models, you can probably pull it off easily with an iterative feedback mechanism with the screen space. Meaning, transform visible voxels from previous scene, see if LOD is satisfied, if not, grab some more voxels from the octree, transform them, and continue until LOD is satisfied.
interestingperson121 2 years ago
Do you think that Nvidia or AMD have understood the power of Sparse Voxel Octree?
i mean, do you think there is some "features" they could implement in their future harwares to make use of sparse voxel more efficient?
Sorry i'm not a pro...
kebyh 2 years ago 3
It will take another 3 to 4 years before Voxel Rendering hits retail. The GPU power will reach the point where you can render voxel at 60fps. But in the year 2010, it's still not possible.
iD0NUTZ 2 years ago
voxels & animation ?
i still think future will be hybrid, subdiv surfaces/displacement-maps
concept of scanned 3d is awesome though - e.g. lightstage
walter0bz 2 years ago
y not ? enemies made of voxels animated?
rutgerhaase 2 years ago
I see a lot of people argue that voxels will not be the future of gaming due to hardware limitations.
But think about how fast our technology is evolving, and has evolved. This just ten years ago, when games like Turok looked "so real", now look at it compared to something of today. And think ten years before that.. Atari..
We'll get there in time. If not on the next generation of consoles, then the one after that, along with VR goggles :-D
jeremysart 2 years ago 2
I've never really understood the practical difference between a voxel technique and modern normal-mapping... Isn't it the same concept?
monsieurouxx 2 years ago
Normal mapping is a geometry approximation which fails with a moving camera, at oblique angles, representing surface concavities, and incorrectly shows sides of surfaces you cannot actually see causing lighting artifacts. Normal mapping is also typically associated with polygons, and polygons have limitations on how many you can push through the hardware.
zelexi 2 years ago 2
SVOs on the other hand are only limited by how much hard drive space you have and the compression algorithms you use to reduce the amount of hard-drive space you use. You can represent a nearly unlimited amount of geometry and geometric complexity.
zelexi 2 years ago 2
I should mention that SVO is an approximation as well. While it doesn't fall prey to many of the typical approximations, if falls prey to another. For one, voxels enumerate a surface at discrete points. For example, a standard definition TV signal is blurry where as a HD signal is sharper and more accurately represents the original scene. As such the resolution of the voxels affects how accurately it represents the original scene.
zelexi 2 years ago 13
A good question along the same vein which really gets to the heart is, what can I do with SVO that I can't do with polygons?
I think that SVO is a simplifying factor. It solves LOD, streaming texturing, streaming geometry, uv unwrapping, etc... all in *one* single algorithm. A giant hammer of which to make game development easier and higher quality. The end result of voxels is what many other algorithms struggle for and what voxels achieves relatively effortlessly.
zelexi 2 years ago 35
I agree, in part. But, there are always going to be issues. I see voxels as an answer to liquid simulations and volumetric effects. As well as surface practuring. A more genuine issues is still the process of creating structured characters. which would require a 'surfacing' effect to the outer voxel points, defining a shell, or even the idea of positive/negative spaces for physics simulations, which might be heavily reliant on a system similar to BSP...
jdzero 1 year ago
@zelexi Don't you still need to unwrap to generate the source artwork used to generate the voxels from?
tehsimo 4 months ago
@tehsimo not when the source artwork is in voxel form. For example, when made in 3d coat.
zelexi 4 months ago
@zelexi They couldnt be dinamically rendered so far surfaces use less voxels than near ones? For example, using the same amount of pixels on screen in voxels. So when you get near, the models always looks "good" and never pixelated, maintaining always the same amount of voxels on screen. Could be that possible?
kitsune0video 5 months ago
@kitsune0video Thats kind of the whole idea :) To only load a trace against the voxels that you can actually see.
zelexi 5 months ago
@zelexi this also true with polygons though. the difference being that mesh detail, texture resolution, texture filtering techiniques and more is also affecting the end result not just output resolution and input object resolution.
koyima 5 months ago
@koyima very true. Triangles aren't perfect either.
zelexi 5 months ago
@zelexi That definition of a normal map is a little false. Normal maps are agnostic of geometry and instead simulates the reaction between surface variation, light source and observer. Normal maps are a an efficient optical illusion, which is true of all good pixel shaders.
Gusdor 1 year ago
@Gusdor I actually entirely disagree with that. Normal mapping defines the normals across a surface. It very much is a geometry approximation. How you use those normals in a shader are irrespective of what the data represents.
zelexi 1 year ago
@zelexi I'm no graphics programmer, but I think Gusdor is technically correct - certainly a height/displacement map would be a geometrical approximation - a normal map is derived from this, but looses certain information, i.e. a normal vector parallel to the surface normal would be assumed to reside within the polygonal plane, whereas in real life it may be displaced above or below the plane. The normal map in this case gives no clue as to the real geometry it was derived from.
antzrhere 8 months ago
@antzrhere Using a height/displacement map to make a normal map is using an approximation of geometry to make another approximation of geometry. Two levels of approximation are still an approximation. The correct way to make a normal map (with fewest levels of approximation, and maximum quality) is to make a very very high poly mesh, and a low poly one. Then project every pixel on the texture from the low poly mesh onto the high poly one.
zelexi 8 months ago
@antzrhere I should also mention that a height/displacement map of the same resolution as a normal map, cannot actually represent all values of a normal map. It can only approximate a normal map. While this is normally good enough of an approximation these days, back in the day when normal maps were first used with very low poly meshes, this was a very horrible approximation.
zelexi 8 months ago
@zelexi I agree,they're both approximations, and I can imagine displacement maps suffer more as the degree of displacement increases with constant resolution.How I see it though,displacement maps could be *crudely* converted back into a polygon mesh (of course, this would only hold true if there were a single topological layer -grooves within grooves cannot be represented with a single displacment map),but a normal map couldn't on it's own.
What were we originally discussing?! :-P
antzrhere 8 months ago
@antzrhere Just because the transform is not reversible, does not mean it isn't a geometry approximation.
zelexi 8 months ago
@zelexi No, your correct.It is a geometric approximation -I suppose geometry as a definition encompasses volume,points,lines AND angles,so I must confess,I was originally wrong.Furthermore an 'approximation' has been defined as:"an inexact number, relationship, or theory that is sufficiently accurate for a specific purpose" - so if it gives a sense of shape (be it believable or not) it is an approximation.In the context of re-constructing volume it is not an approximation,but thats irrelevant...
antzrhere 8 months ago
@zelexi Just a quick question -with 100% screen coverage(but not completely zoomed in),what sort of performance do you get on GPU (million rays/second)? About a year ago I tried to implement something similar -it worked nicely - ray-guided streaming,compression etc. but as it was in software,on a 3.9Ghz i7 I could only get 16-20Mrays/sec with 100% coverage - max 30M completely zoomed in. I know Nvidia SVO managed 55-60M and peaked at 110mps inc.filtering.Seems CPUs just cannot compete.
antzrhere 8 months ago
@antzrhere With that demo and a GTX280, about 60M. With a GTX580, about 150M. However, the hardware is capable of better ;)
zelexi 8 months ago
how many voxels are in the model?
DanFrederiksen 2 years ago
35,184,372,088,832 in total voxel space. The model is scaled to take up almost the entire space. There is still internal cavities that is collapsed and some outside space, however the total number is still staggering.
zelexi 2 years ago
that sounds like complete bullshit because how could you possibly have that in memory..
how many voxels are in the model? actually
DanFrederiksen 2 years ago
its not in memory, its streamed in from disk.
zelexi 2 years ago
you really don't know how it works do you?
even on disk it would be more than 100TB
DanFrederiksen 2 years ago
Note that I said "There is still internal cavities that is collapsed and some outside space" This probably reduces it by a few orders of magnitude, as well as there is some simple compression going on. I would recommend reading the paper. That large number that I stated is "total voxel *space*" not total number of voxels used by the model. That is a much harder number to determine without specifically implementing code to measure.
zelexi 2 years ago
Btw, I should know how it works ;) I wrote it! :) On disk it was boiled down to a couple gigs IIRC.
zelexi 2 years ago
you should know yes :) and if you don't know, why should I read your paper... : )
I recommend you do a count of the voxels in the model, seems a pertinent bit of information in evaluating voxel rendering.
it seems that a 2gig model of a single character in datastructure that's impossible to animate, means that voxel rendering is not of interest and never will be. would you agree?
DanFrederiksen 2 years ago
I simply don't have the time right now to write that code and test it out. Perhaps I will in the future when it matters enough to me to do so.
zelexi 2 years ago
It is not impossible to animate. Just there there are better things to do with your time for x720/ps4. You may see animated voxels as I have seen research in this area. The method that I propose is to render the static objects with voxels (namely the world) and the dynamic objects still as triangles. Which makes sense as if you look at games today the characters look immaculate but the world can use some serious improvement. This gap has been closing and voxels is one possible end to that.
zelexi 2 years ago
The 2gig level doesn't do FFT or Entropy encoding, so there is about an order of magnitude possible improvement. 2gig down to 200mb. For example, idTech5 when uncompressed textures used up on the order of 200gb for a 1 level (of which they have many). This compresses down to fit onto a few dvds. The same is likely to happen here, except that here we will have to fit onto a few blue-rays. So this technology is very feasible and practical for next-generation hardware I assure you.
zelexi 2 years ago
Seriously, I answer many of these questions in my presentation. I'd highly recommend taking a peak at it.
zelexi 2 years ago
v-sync much?
ilam3d 2 years ago
It runs only 60fps? Is there any framerate limiter? Because 60fps on only one model using a fairly new video card is not a good thing.
innonni 2 years ago
this is a tech demo. The real thing is still a few years away, by then Intel's Larabee will be out and maybe the entire graphics industry will recieve another overhaul similar to what unified stream processors did to the old vertex/shader pipelines.
dflynchimp 2 years ago
when this model contains over a million polygons, it is actually an amazing thing!
EricTheRed03 2 years ago 2
60fps for millions of polygons is not bad at all.
DanoruX 2 years ago 7
It's 7 millions of poylgons rendered in real time.
...
Fritzoune 2 years ago 3
Not really, 60fps for this is fking amazing. Have you ever tried coding voxels. Performance is the craps on them. This is impressive, and your forgetting the common scale, when a game runs at 60fps *only, pfft, that dosnt mean 2 objects will be 30fps, often a super simple game example with a complex engine can render a cube at 800fps. And u think omg only 800fps for nothing onthe screen thats bad. Yet you can add 10,000 objects and it still will maintain 40fps or so.
IMattNovakI 2 years ago
that model is comprised entirely of voxels. Not a few polygons with a bunch of shaders and crap.
sc0pl355 2 years ago
Real-time rendered? Really nice
DigitalWebOwner 2 years ago
whatever the case it still looks awesome and i can't wait until technology starts being used games.
J0RDSKI 2 years ago
And in fact, it's polygonal meshs that are converted in voxel models! ;)
(Even the opposite seems to me very difficult!! :/ )
Nival39 2 years ago
But what about animating it?
CodyRicheson 2 years ago 2
this is id software dude, sure they wont show stuff like this without knowing they will find a way to use it!
Capeau 2 years ago
Absolutely they show stuff like this without application in their mind, because they know that a lot of people think like you, and think thing like "such a great man! he has invented the future of video-game!".
But in fact: no.
Voxel is unusable (through currentl real-time technologies) for advanced animations and interactions.
Eventually usable for really static environnement (allow very detailed things with easy advanced lightning). Currently noway for anythng else.
Nival39 2 years ago
With currently-known real-time technologies: simply no way (but whole-model rotation and translation ;) ).
Nival39 2 years ago
How many light sources? Seems brightest source is upper-left, yet shadow is lower-left. If there was also a brighter upper-right source, this could make sense... Otherwise, I'm scratching my head in confusion.
MrKlorox 2 years ago
Could be two. The one to the right is at a lower angle so it doesn't shade the side of the body on the right hand side.
In all honesty I don't know what the point is in this Voxel technology unless it actually offers possible improvements. With polygons catching up in terms of detail, what's the point in researching Voxels?
Project Offset looks very much like this, by the way.
3kliksphilip 2 years ago
project offset does in no way look as detailed as this and voxels are easier to handle if you're going for verry high detail.
Capeau 2 years ago 2
Would be cool if a Computer had Voxel and Polygon-specific rendering at the same time, though. Voxels can, potentially, change gaming future as we expect it?
MettallicDesease 2 years ago
why is the only time they rotate the model when it is super zoomed in? thats a pretty slimy trick
illustriouschin 2 years ago
zoomed in and rotating is technically a worse situation, because the more zoomed in you are, more of tree is not loaded in when you rotate around seeing pieces of the model you have never seen causing an even worse immediate loading situation.
zelexi 2 years ago
Cryengine 2 uses voxels for some geometry. As the terrain is heightmapped they use voxels to make stuff like cliffs, caves and rocks in their terrain as the heightmapped surface cant.
friggindoc 2 years ago 2
voxels are used as modelling tool in the case you described, but in the end it still gets converted to polygons, at least that's what i was told.
but i heard rumors about voxel-techniques being integrated into cryengine 3, although in the end everything gets converted again.
azziplxx 2 years ago
Absolutely not.
Voxel is a kind of 3D models which can be directly display like voxel, in fact it's even the interest!! because this kind of display have advantages, like very high performance for very detailed models (thanks to a built-in adaptative level of detail, automaticly combining both 3D model and texture-resolution adaptation in the same time) , and easily lightning (with dynamic shadow, etc).
Prob: currently no way for advanced animation and interaction. So static environnement only.
Nival39 2 years ago
voxel dont get converted.. often they are raytraced. you got that wrong ;)
ecreif 2 years ago 3
my kids will play this )))))))))))))))
plandercozeron 2 years ago 21
ah, the good old voxel zooming game, eh?
abdigon 2 years ago 3
Can the voxels be dynamically lit and shadowed? Cast individual shadows etc?
scorpius420 2 years ago 15
@scorpius420, yes, because each individual voxel contains light and color information.
CodyRicheson 10 months ago
Is there any chance os voxels like this being used in games again? I still remember Outcast and Commanche! :(
scorpius420 2 years ago 15
comments on this thread are silly; voxels could perfectly be used for all 3d entities (terrains, characters, etc), the problem is current hardware is designed with polygons in mind; technology created with voxel graphics acceleration exists, namely in the medical field for TAC scans visualization and whatnot, we just need a change of paradigm, maybe in the form of a gaming console focused on voxels... ...
Nouseeker 2 years ago 32
True, there are some places that would make it for voxels a problem in games. However that's why I think the future is gonna be more of a hybrid, Polygons still will hang around as well as mainly software render/programing/shaders taking over that will be accelerated weather it be voxels or whatnot.
And having a good voxel modeling program would just be something like Zbrush but using voxels. And polygon -> voxel is always a option too, or in some cases maybe voxel->poly for destruction's.
IMattNovakI 2 years ago 12
@Nouseeker problem with voxels is although they make beautiful rigid still images. hardware just cant cope with that level of detail when you introduced movement.
jakspyder 1 year ago
@Nouseeker But how can you calculate the matrix deformations on millions of voxels at an acceptable realtime performance?
The voxel technique seems great for static meshes. But for anything that needs to deform, I think traditional Polygonal systems plus hardware tessellation are the better choice for the foreseeable future. (until we have hardware that can deform massive voxel datasets.)
SSCrow 1 year ago
@Nouseeker You mean a... Vox Box? ;D
ZoomingDakota 1 year ago
@Nouseeker Voxels can't be animated. At least not animated well. Show me one example of a voxel character that is not animated with frame-by-frame. Basically, voxels mostly suck for several reasons -- A) No way to animate a character well -- B) Filesize grows exponentially larger as you increase detail -- C) Difficult to texture (requires paint directly on model, no uv / unwrap)
mcapplbee 1 year ago
@Nouseeker
Not going to happen, all the technical stuff aside, we've just perfected the polygonal techniques way too far to consider a switch, the amount of time and effort required to change factors for the simple benefit of seeing more detail on extreme close-ups is just not worth considering that in the end it is the overall experience of the game that counts, not what the character looks like if we put their fingerprints up to a 200x zoom.
Shitsocash 1 year ago
@Shitsocash Thats a valid criticism for characters. However, it this was world geometry and not a character thats entirely different story. And really, the tech is for world geometry, and not really characters all that much. I only used characters for the siggraph demo cause thats what I had.
zelexi 1 year ago
@Shitsocash This is the same logic that is preventing us from colonizing and making material use of other planetary bodies. Because we choose to take the least short term cost for the most conservative reward regardless of how much it will cost us in the long run. Hopefully Asteroid Toutatis will cure our lack of foresight!
illustriouschin 1 year ago
@Nouseeker It's not just a rendering problem. There's also the issue of animation. While voxel solids can be animated perfectly well, the tech doesn't lend itself well to the traditionally mesh-based deformations needed for soft bodied, flesh, cloth, etc. Until this is solved, I think we're pretty much just going to see hybrid engines.
GotNextVideo 1 year ago
@GotNextVideo
I believe animation is still possible, I remember lattice skinning being mentioned somewhere. Atomontage has an example of a hybrid polygon voxel engine.
1776Phoenix1986 10 months ago
Looks like a standard polygonal model that was made with extruding a height map instead of with a modeler.
What am i missing here.
Outlander11 2 years ago
The fact that it's in real time.
This demo uses a volumetric octree, or a representation of space in which each node of the tree is divided up into eight parts, and as you descend down the tree you get a more refined approximation of the original, high-resolution source material.
If you tried doing this brute-force, using the original polygonal mesh, it wouldn't be real time. It would be a slideshow.
RoboDonut 2 years ago 5
Thanks for that great explanation! This is very impressive. Can the same be done with textures? Or will it ever be the case?
shletten 2 years ago
Absolutely, textures are fine. The source model (Imrod) had no uv information so I couldn't put textures on it.
zelexi 2 years ago
I see, but that's not what I meant by that. RoboDonut said that this voxel model was divided in 8 parts and the more you got closer to it, the more defined it became. In actual games and movies, the closer you get to a texture the more you see the pixels and it eventually becomes very blurry. Can a similar LoD approach be done with textures? That would be very cool if one day it'd be the case.
shletten 2 years ago 14
yup :) this is kind of built into the technique, as there really aren't textures anymore. Each voxel has both position data and texture data inside it. Similarly, there is no need for normal maps for geometry approximation anymore as we can represent this geometry directly.
zelexi 2 years ago
That has already been done (Paged virtual texturing, as seen in Quake Wars and rage), but it would actually be moving backwards. With THIS technology there would be no more need for textures at all - all detail would be (or at least COULD be) modeled
Wobbothe3rd 2 years ago 13
That is incorrect.
Vamavid 2 years ago
Im not so sure that textures are done with, you obviously need some kind of colour(texture) information or all youll get is grey or single toned models, but thats why I think procedural generation for both texture information and these voxels would be the future, assuming voxels are able to be animated, if not then that obviously limits Voxel use to static geometry, which is becoming less and less used in todays games
Atomizer74 2 years ago 13
Wow, it's like real plastic. Amazing.
LHCLEGION 2 years ago
Voxel is cool for farther Terrains in RPGs. Crysis uses Voxel to render them. As long as you don't get near to it, it looks like real.
TheRealNici 2 years ago
in crysis voxels are used only for caves and complex terrain, that u cant make using polygonal terrain because it can be edited only on z axis
piroman665 2 years ago
awesome
P4N1CM0DE 2 years ago
It's a pity that I paid $20 for this video on Siggraph Encore ;)
kixorz 2 years ago 7
2 jeffgooloolie: You can't on the current technology.
kixorz 2 years ago
You can on PC. single 280gtx ran this demo at 60fps 1200p
Tombes123456 2 years ago
but not animated!
make him walk and it wont work anymore.
They are only going to use this tech for the enviroment. Characters will still be made out of polys. Unless the next gen hardware gets better then everyone in the gaming industry thinks it will.
Capeau 2 years ago
Yeah, but you -cant- animate it, correct?
jeffgooloolie 2 years ago 4
You can't on current technology.
kixorz 2 years ago 2
You can on PC. single 280gtx ran this demo at 60fps 1200p
Tombes123456 2 years ago
but not animated!
make him walk and it wont work anymore.
They are only going to use this tech for the enviroment. Characters will still be made out of polys. Unless the next gen hardware gets better then everyone in the gaming industry thinks it will.
Capeau 2 years ago
It would take an impractical amount of data to animate any high-resolution voxel set.
This is probably going to replace heightmaps soon, but not skeletal animation.
cyborgtroy 2 years ago 5
Can someone I.D. the track being played in this vid?.
MarlonBrando341 3 years ago
OMG wow.
Look at the detail!
EricTough 3 years ago