Hello, yes my title is sort of provocating... but let me explain. Some people will think what the hell is he expecting? I am very perfectionistic, if I had the choice between finishing a game that doesn't look good or don't start with developent at all, I would choose the latter one... I am not new to the game dev, but I always end up in terminating a project due to lack of digital content and animation. This is a rather long question but I will summarize it at this point. Please remember I am talking about real-time, in-game things with hardware some years in the future (since programming my game will take a while): 1)What about realistic motions like in state-of-the-art games? Is there any way to get something close for a single developer? Most hobby/OpenSource games out there, including my ones, look like robot invasion, at best... 2) Is there a better software for clothing simulation than DAZ Studio with the "Dynamic Clothing" plugin? I mean last time I was using it, only two dynamic clothing bundles were available. So you can do nothing productive with it, even if the idea itself is almost perfect. 3) Is there any good terrain rendering software out there where you only specify global aspects, like heightmap and areas carrying information about wood, snow, water, etc. But then the software should use some fractal-algorithms to interpolate all the details from my global specification...? This is possible, just before I start to do this myself I want to know if there exists anything. 4) Is there any good software out there, where you can build realistic architecture with the least design effort? Again this should work in the way that I specify global constraints and the engine computes all the details and variations using fractals/heuristics, or whatever. 5) I think this last question is useless... Is there any Software that now can apply realistic physic effects to my previously created digital content and save it in a format I can use in my game? I mean the DAZ Clothing is rather useless if I can't reproduce it in real-time. -------------- The long explaination I know tools like Maya, 3DS, DAZ3D Studio, Blender, Poser and all that kind of stuff. But none of them seems to give a single person any chance to create competive digital content in reasonable time. What I mean is simply that it will probably take years if not forever to create good characters and animation, the environment (flowers, buildings, etc) for a single developer. Even if DAZ Studio or Poser are more the ones I am searching for, they still lack many important features and not mention, they only provide good support for character animation but not for environment setup. When I want to create a game, I have an idea in my mind, a vision. If you look at games like CoD 6, Uncharted 2, Crysis, etc. you will most likely see that they all look kind of "similar". I don't say that they look the same. But after all most of their renderings are generic, meaning that they are not related to the story or the game itself and could be reused for any other shooter with realistic environments. ------ Part I: The Environment When you look at algorithms like fractals, Mandelbrot, perlin noise and whatever (I also used them in my own terrain renderings) it seems that at least to some degree it should possible to provide realistic rendering without designing all those details by hand. So one could age metal, stones and whatever, by applying stains or growing plants on them, etc. An engine could even burn all those things because they way it is "meant to be burned" stays the same, regardless of the game itself. So why isn't there any engine out there where you can create such realistic environment without having to model it in every detail? It should definitely be enough to specify the heigthmap and area-based content information, like snow, water, flowers, wood, or paths where the player should be able to walk, etc. Based on the heightmap and area-info it is really not that hard to generate realistic looking terrain, since I wrote such an engine myself. But of course it was still lacking many important features... With this little design effort the engine could already provide interactions like footsteps in snow, burning wood, etc... ------- Part II: Buildings The next thing are buildings. If we build houses stone by stone, why don't we do it in digital content creation? This would automatically bring a realistic look and feel to all buildings and most importantly, they would be all fully destructable, because the engine now knows exactly how it can destroy any part of a building. Is any software out there where designing realistic (preferable destructable) buildings is not a life's work? I mean I can say that with every software I know so far it would definitely be faster for me to build the building myself in reality and take some nice photos ;-). ------- Part III: Character Animation and Clothing The most complicated thing is animation. Of course Poser and DAZ are great tools for guys like me but they lack important features to be useable. The motion of bodies often comes with "artifacts", for example if you rotate a forearm, it will look like you were twisting a piece of rubber. Not to mention the clothing itself. DAZ provides a new feature called "Dynamic Clothing" which would be perfect, just that there are no good clothes supporting it ;-), which renders it rather useless. You can almost forget about all other approaches like static cloths, that is just garbage to me, unless you are rendering still images. But thats not all about animation. What about realistic motions like in state-of-the-art games? Is there any way to get something close for a single developer? Here again I have many thoughts about automating all this crappy stuff. For example the PhysX engine theoretically provides all the things necessary to simulate clothes and thus hair, flesh, etc in real-time. The motions itself could be precomputed by using realistic reference models for a huming being (muscles, skeleton, flesh, etc.). This way a not too complicated artificial intelligence could pre-compute a realistic motion without me having to specify every animation step. Instead I would tell the engine where a character is located and what you should grap, for example. Then the AI could compute the motion to grap that thing and we are done... As I am a very experienced programmer in other areas, I am sure that none of the techniques I described is impossible. So my question is why isn't there iny engine or tool set supporting them? Since it would definitely make game development a lot easier if not possible at all for me. best regards chris

thanks for your reply.

I know you can compare them with blockbusters ;-). But if you already do this comparison I have another one.

The concepts above are not meant to really compete with cutting-edge design. It is just the cheap way of creating realistic environments even if they will look artifical, but that isn't so important if they look good and the game is OpenSource or a hobby production. I would compare it with those movies, produced with a capital of some millions and still competing or even beating blockbusters in the story and acting quality.

I don't even imagine any engine producing cutting-edge design only given global constraints and I think it will still take years if not a decade until an engine can automatically compute such "interpolated" renderings in real time. But I am pretty sure that you can get close to it with current hardware, at least close enough to let the end-user overlook the not 100% cutting-edge graphic because of the great story and gameplay ;-).

regards
chris

Quote:Original post by Konidias
A lot of the stuff you've mentioned would bring even the most expensive personal computers to their knees.

I think this depends on how far you will go. If you look at the latest SDK demos from NVIDIA and ATI you already see amazing content and also real-time cloth simulation being "disturbed" by real-time force fields like wind or rigid bodies. You can even break any kind of material and stretch cloth until they break. And it runs with 150 fps on my GTX 260.

So at least the clothing part is already reality, but there is no useful tool to make it accessible for game development, at least not if you don't have millions of capital to waste.

Terrain and Building generation is also depend on how far you want to go. Building could carry LODs, so only buildings in the direct environment will be rendered as huge collection of rigid bodies. If the building is some distance away, it will be replaced by some optimized/static version. Also it is clear that such generated content will look artifical and generic in some way, but at least it will look good!
And again for example SpeedTree and the one NVIDIA Demo already provide the real-time effects you mention, like grass waving in the wind as well as entire woods waving in a breeze... This is just a question about optimization and reducing complexity to a degree that still looks "realistic" but can be computed on modern hardware. The problem is that SpeedTree for example costs about 5000$ per title. And you would only get the trees with it ;-). There we go...

regards
chris

Quote:Original post by swiftcoder
This is unfortunately the reality: at the present time, procedural content generation is generally *more* expensive than an art department.

Well but the good aspect would be that it would only be more expensive ONCE, since many games could now use this generation.

1) The industry uses rotoscoping, taking the movements of live actors. They often use expensive equipment for this, but it is not strictly necessary. You can get by with camera footage, ideally from two angles, plus a lot of work making your animation match the footage.
You should be able to do pretty good animations like that. The catch is that it is a lot of work, more than most hobbyists are prepared to do. Hence the 'robot invasion'.

3) Bryce generates landscapes using fractals. You can download version 5.5 for free.