I recently made a rather long but random post about unlimited detail for real-time rendering. The post regards the awesome claims of a company called Euclideon or Unlimited Detail, which is lead by Bruce Dell, CEO.
Here is also what I think is an older website from the Euclideon group I found,
I am pretty convinced from watching the live demo on the laptop that the company is serious about its unlimited detail claims. Many people have gone to say this is a hoax, saying the tech demo is a pre-rendered scene, or that it is unable to have moving objects, or that it is just impossible because it would defy so much of what we know about 3D rendering technology already. Notch, Markus Persson, the famous creator of Minecraft has even come out with some statements that he believes this is just a hoax.
Many people are considering it is voxel based technology. Minecraft in essence is like a voxel based game although I do not think it actually uses voxels it is still a good example. Minecraft creates huge random virtual worlds with big huge chunky blocks. Supposedly a minecraft world can be larger than our own real world earth by a fairly large amount, but the world looks fairly low detail, and chunks of the world are being loaded in and out as needed depending on what the player has within their line of sight.. A voxel would be like a cube of space or virtual matter, and all objects in the world are made of these cubes, which are in a way similar to what Euclideon talks about with virtual atoms.
Voxels have many limitations though with how many cubes can be on the screen at one time, and limitations with animation. One talented fellow seems to have figured out a great way to overcome the issues with animation which he covers in this video here which is exciting news.
I personally don’t believe it is a hoax that Euclideon has figured out something short of amazing and revolutionary, and I would be terribly sad if it is. First of all, knowing quite a bit about special effects and pre-rendered animation, I can tell you there is no indication to me that the graphics on the laptop are faked. It appears to me to be an actual genuine demo of the engine running in real time.
This technology Euclideon has shared represents a lot more than just a revolution in games, it is a revolution in all 3D visualization across the board if it can gain some ground. This could greatly improve the field of not just visual effects in the long run, but product design, medical imaging, and more.
Many artists are limited by the constraints of using polygons to represent objects. When you first start working heavily in 3D, this is seriously no joke in that you are required to learn and follow a lot of rules. If you have never created a 3D model then you have no idea how non-intuitive it can be for a typical visual artist.
An artist must first learn a host of technical skills in order to achieve good looking results. An artist cannot rely or take advantage of many of their natural design skills until they have mastered some pretty complex tools and technical theories first. When you pick up a piece of clay to create a character, you can shape it and mold it however you choose and very little technical skill is required to make something cool fairly quick. Need to add an arm? Just shape one and blend it together with your characters body. The process of adding and subtracting is fairly intuitive.
Attaching an arm to a 3D character is a bit more of a complex process. Working with polygons requires a lot of time and practice because there are many rules regarding topology and optimization that make no sense at all to a beginner. For starters, the more detailed and realistic you try to make it look, the slower your computer will run. Working on high resolution characters for film effects often will slow my computer down or cause it to crash even if I push it too hard. A great deal of pre-planning is involved as well if a character is to be animated in particular ways, because polygons do strange things if they haven’t been properly structured.
Detail optimization is always a major issue. Sometimes separate models must be made if a model is going to have a really closeup shot, because only so much detail can be on the screen at once before operations begin to slow to a crawl. To save time in the long run, an artist might make a character model with no surface details like warts or pores in the skin, and hair is a maybe just painted on as a texture. This model is used for far away or medium camera perspectives. Then if a closeup of the head is required, a highly detailed duplicate head will be made, with pores and wrinkles and bumps, individually rendered hairs, the works. That detailed head may not have a body though and just be a floating head, the computer can devote more resources to making the head look more detailed if it doesn’t have to worry about generating the image of the body as well.
Not just in the modeling phase alone but also in the texture creation phase, artists will encounter extremely counter intuitive tools which are required to learned in order to achieve professional results. The process of UV mapping a 3D object for example is certainly tedious and sometimes a bit odd, you must essentially flatten your model on an XY plane so you can make and apply a user editable texture map. The newer modeling tools are making the process easier, and the ability with painting models directly in mudbox or zbrush even makes the entire process of dealing with UV’s easier and nearly invisible. However in some cases it is still required to layout a UV map if you want more control over your textures.
There has been an amazing revolution in the way models are created, and that is with the introduction of 3D sculpting software and for the first time we are seeing an intuitive way for artists to create without having to worry so much about technical issues concerning polygons. Of course the technical limitations of polygons are still present, but they are actually somewhat avoidable now in some cases. The first I ever heard of 3D sculpting was regarding a software called amorphium.
This application allowed a 3D artist to use sculpting like tools for the first time to shape models in a similar fashion to working with clay. Amorphium was a breakthrough, but unfortunately it was very limited and seemed to be difficult to achieve precise results. To be honest, I didn’t think the software was going to go anywhere at the time, it seemed like a novelty rather than a tool. It just wasn’t precise enough to create a high level of detail.
Years later a company called Pixologic begins picking up steam with its product Zbrush. Autodesk comes along with mudbox. These two apps take the concept of sculpting like clay with 3D models to a whole new level by allowing a level of precision and detail with a more advanced toolset. They have revolutionized the 3D modeling industry, allowing artists to use much more intuitive and natural tools to design complex and highly detailed models.
However there is a growing issue with the sculpting programs. They are also becoming increasingly complex with the new versions every year. The tools and workflows become more advanced and complicated. For example, you can sculpt a model now and get a shape you like, then re-topologize the surface so that the polygons are layed out more efficiently for animation. This process is still fairly boring, tedious, and very time consuming.
I think it is great that the developers are constantly adding new useful features, but the process of working with polygons as you start to work in higher detail is still a very tricky and complex one. Models can now easily reach the level of millions of polygons, which in the recent past was unheard of. The more detail you add though, the slower your computer will run and the higher the chance you will run into memory crashing.
Now lets just say that a technology like Euclideon’s Unlimited Detail using point cloud rendering, or graphical ‘atoms’ you could say, was to be used as the core rendering engine for future sculpting software used to create 3D objects by artist. An artist could truly work with a surface that emulates clay and allows them to arbitrarily modify the sculpture without worrying about how the surface flows. Surface topology becomes a thing of the past. Not only that, but there would be no slow down as more and more detail is added to the character.
Need a little bit more mass on the nose? Just add a few more atoms here or there, color them and shape them as necessary. The whole model structure truly becomes organic, just like real world clay which is made of tiny atoms.
The artist would no longer need to consider the topology of the surface, which is required to make a well animated character and forces an artist to go about the work in a restrictive and non-intuitive manner. Not sure what surface topology is? It is basically a term used to describe the way in which the polygons that makeup the surface are arranged and connected together at the edges. If the polygons are connected in certain configurations it can cause issues and artifacts when the model is later animated or visually rendered for a final image, while other configurations will produce better results.
Artists currently must take care with deciding their topology before making a model very detailed since topology cannot easily be changed usually after the fact. If they skip this process their model will potentially look terrible when animated. If you are using a standard 3D application, or a dedicated 3D sculpting application, you still must make careful choices about your polygon topology. Zbrush is getting better tools to deal with topology with each release, but they still have not fully automated or hidden the process, and it is still a consideration for the artist.
Another interesting thing about polygons is that models created with them are typically hollow, so there is no skin layer, fat layer, muscle tissue, bone, nervous system, etc. Behind the skin of a 3D character is nothing but empty space, unless an artist models the insides parts there is nothing there. If you want to have a characters arm cut off, you need to make two versions of your model, one with the arm connected, and one where the arm is severed as a separate object and you have created a ‘cap’ for the severed arm and the arm stump. With polygons the topology cannot easily be changed, so you cannot easily create a severing effect.
With ‘graphical atoms’ it is possible the object can be volumetric, meaning it might have an inside which is solid rather than being hollow by default. If that is the case, you could potentially slice off a characters arm at any position you want and not have to have any pre-planning in regards to topology. Unfortunately it is too early to tell from the demo if volumetric objects are possible, and also what sort of physics can be accomplished.
For example, if you had someones arm cut off, you would want an appropriate physical reaction to occur with the severed part. It would be a combination of soft body physics and rigid body physics to achieve the proper motion. There would also be a sort of liquid or particle system to simulate blood. The soft body being the fleshy parts of the arm which would jiggle a bit, and the bone would be the harder structure the flesh is attached to would be a rigid physical object.
At the very least, a system like what Lucas Arts created for the recent star wars the force unleashed games developed would be desirable. Lucas Arts created a technology they call digital molecular matter.
It was a system to simulate certain kinds of advance physics effects. Like a piece of metal would bend like metal when great force was exerted on it, or a piece of wood would splinter into shards if struck hard enough. The tech demo’s looking amazing at the time, but the implementation of the technology in the game was not so amazing. They had to use the effect very sparingly, and ended up barely using it at all. This is because it requires a lot of processor power to simulate polygons to act like matter.
Perhaps the engine Euclideon is working on would be far better suited for a system like DMM, considering it is already creating objects made of virtual atoms the realism could be even greater. Lucas Arts DMM, is still limited to using polygon shapes which most likely keep the same topology. They are splitting pieces of wood into smaller pieces, but they are probably not doing so by changing the topology. They are probably just swapping out the model with a more complex model made of pre created chunks. The Lucas Arts DMM is calculating a lot more variables such as the impact position on an object, the impacting objects force and mass.
However I think in essence it is doing the same thing with ‘model swapping’ that is shown in the beginning of the demo. When the wood is being splintered, it is not creating 100% unique shards. Some of the splintered pieces look like they are duplicates of other pieces. With unlimited detail though, you really could make completely unique shards if you could find a way to program the physical rules into objects.
Just like Lucas’s DMM, you would need to figure out the speed and mass of objects, but also create a sort of library of known material types that would have certain ways of reacting to great force. Wood will bend and eventually snap, glass will not bend but instead would shatter, stone would crumble and at times shatter as well. This is not something the engine has yet though obviously as they have been focused up until this point on achieving the Unlimited Detail capability, but I do recall Bruce Dell mentioning in a video that it would be possible to do physics.
In the last post I brought up Atomontage Engine. This is a voxel based engine which in some ways is similar but wholly different it seems than what Euclideon is offering. I recommend you look at the videos released by the developer though, as it is a very interesting engine..
You can see that in regards to atomontage being used as a game engine, the developer ( a one man operation btw ) is showing off some interesting features. The car in his video can leave tracks in the sand in real time, deforming the terrain. Also, he even shows how he can dig in the sand to reveal hidden objects such as smaller rocks within the volume of the ground.
Building destruction would be far more realistic than games such as battlefield 3 which is currently one of the leaders with this sort of feature. Battlefield 3 is breaking objects into pre-defined chunks though and not creating completely unique chunks. This is because the variety of topology changes that occurs in the real world when a building is destroyed or a hole punched into a wall by an explosion cannot be easily generated on the fly. Having the computer switch from a solid wall, to a wall that is fractured requires swapping between two polygon models, a solid one, and a bunch of broken pieces.
In the atomontage demos, we can see that he can cut into a buildings wall, cutting through the outer layer of plaster, into the underlying brick. The holes he creates look more natural in some cases than games like battlefield, but it is lacking the physics simulation. For example, if he cuts away the base of a wall, the wall does not fall over or break apart like it would in the real world. It just floats in midair, so the developer obviously has a lot of work to do on the physics side of his engine. He does show that he has implemented physics on the vehicle in his demo, so chances are he is working on figuring out how to get the voxels to work similar to the way Lucas Arts digital molecular matter works.
As cool as the atomontage engine demo videos are, there is one major difference. Unlimited detail.. The atomontage engine has limits, and although it can create some highly impressive results already and some very detailed scenes, it probably cannot achieve unlimited detail.
Unlimited detail just sounds ridiculous to be able to achieve at this point on a consumer machine, like something that we would need advanced quantum computers to be able to even consider, but I would not be surprised that Euclideon has just discovered a different way of doing things, one which is obviously much more effective.
Technically speaking, the detail is not truly unlimited for the average user in fact, because detail is physically going to be dictated by the size of your viewing device and the size of the pixels in your device.
The detail is intrinsically limited to the maximum size your computer monitor. It would be really cool if the technology is also unlimited in its resolution scalability, meaning the larger the monitor and smaller the pixels the clearer and more detailed the image becomes without any cost to processing or speed. I mean, the software might run nicely on a small screen, but what about a larger screen where more detail is on the screen at the same time?
Let’s consider an extreme example. Say for example if you went ahead and built some crazy virtual reality ‘cave’ room in your house where the walls are all super flat OLED screen material. The resolution of your cave would be something ridiculous, lets just say for one wall alone it turns out to be something like 1920×1280 times 40. That equals a total resolution of 76800 x 51200. That is huge. If the Unlimited Detail software can still run super fast even at that screen resolution then it is really beyond ridiculous and Euclideon have reached a level that Unreal Tournament players at least would consider, ‘Godlike’.
Lets not forget, companies and artists have invested tons of money into polygon tools and training to use them though. Polygons are BIG business, and are actually a huge part in generating more money than just about anything in the world right now if you consider their widespread use in video games, and visual effects for film. There are multiple industries that would be affected by this in a big way, in a good way for the artists, but a bad way for the companies which have stuck with polygons as the primary tool.
So just like with oil companies and sustainable energy, it is going to be slow before the world adopts the new way of doing things. I think though it is possible that the benefits will be so great it will ultimately be unavoidable. It seems perfectly logical to me at this point, that a system based on atoms would be far superior to polygons.
I hope in the future artists will be able to interact with virtual matter in a way that is far more realistic and intuitive if tools are built with the unlimited detail engine. Though with Euclideon being so secretive and hermit like, it is scary to think they might just disappear again like they did after their first announcement of the technology and perhaps never return.