Real time

[rahulkul2003]

Hi all,

As this site is dedicated towards cg architecture and other related stuff and i am into game art person so i handle some game engine like crytek and udk.

 

While work on games part i saw lot of people have upload real time architecture walk through on youtube done in game engines.

 

link

 

So can anyone tell me is real time is the future of 3d architecture visualization.Do really companies demand for this kind of stuff or they still focus on V-ray quality.

 

Thanks

[JonRashid]

I personally think there will be more demand for this type of technology. It allows for realtime amendments, mutiperson usage and also works over the net if using the right engine. The detail, lighting etc will soon be of a similar quality to pre-rendered http://www.break.com/index/video-game-technology-allows-unlimited-detail-2116780 so why would it not be used where appropriate. Printed images will still play a part but I feel realtime has a serious role to play. Also augmented reality improvements alreadys allow for the placement of a building on its actual site. This could be invaluable for planning, renovations and heritage work.

[cgfx]

i have yet to try the implementation of game engines into arc viz - closest i did would be a level design i did using half-life engine which goes way back

 

it would certainly be nice if such needs/demands blooms

 

any showcase/sample of augmented reality used in arc viz? i saw the honda advertising campaign using augmented reality, which is rather cool. but i have yet to see actual usage of such tech in architecture

[JonRashid]

http://www.unrealengine.com/showcase/visualization/cowboys_stadium/

http://forums.epicgames.com/archive/index.php/t-745454.html

 

I am currently working on projects that involve historic sites the user can walk around virtually in stereoscopic. Even when shown on a flat screen rather than a cave system it has an immersive feel. This allows for a feel of the space involved as well as allowing the user to have complete freedom of movement and field of view. When more end users experience this freedom and see how it facilitates some design and construction decisions in a format that is intuitive rather than requiring interpretation I think that will drive the market. It may be a games engine that assists in this as their combined R&D budget far outweighs archvis development. Autodesk seem to be striving for realtime results with every release and as graphic card capabilities increase towards this goal it is surely inevitable. I think there will be a period of time where compromises in image quality may have to be made but if you have a look at NVidia's own development kit for realtime presentations some remarkable results can be achieved.

It will be resisted as almost every advancement seems to be. People don't like change especially to their working pipeline, so I think the uptake will be driven, initially by end users.

[Jonathan]

John,

Im also involved in heritage visualisation and have spent some consideable time with photgrammetry applications. I know this is an area heritage is very interested in but is it possible to intergrate relatively high poy models and texture maps derived from point cloud generated systems into realtime?

On the architectural note, I personally think realtime is particularly relevant to architectural proposals, however, when architects get to know about this they will want your work yesterday and think of all the ammendments!

[JonRashid]

Haha they already want it yesterday. We are currently experimenting with point cloud conversion. As using 2 kinect's we can generate a textured point cloud. The next step is converting this into an optimised polygon model which we will have to write ourselves. Autodesk have already got tools to manipulate point clouds in max but they are limited.

Once we have something more concrete in that area I will let you know

[Jonathan]

...you could try Meshlab for point cloud conversion to mesh. Its also really useful for cleaning, filtering, merging and re-meshing dirty scans, then a simple retopo, with normals and diffuse mapping high poly to low via render to texture in Max, its is a process that has served me well. Texture information is a little more complex but I think the guys at Meshlab are working on something. Although still in animation rendering, some of my results can be found here: www.tidalpoints.wordpress.com (the shoe model for example is less than 2 hundred polys)

[JonRashid]

thx for meshlab tip. I know my programmers will still persevere however it might be of assistance to them and would certainly give me something to start testing my kinect point clouds with

[Jonathan]

Dont know if you've seen this yet but I nearly wet my pants!!

[JonRashid]

That's 2 of us then. I wonder if the dynamic update will be part of sdk functionality but either way it brings great functionality for so little cost.

[Scott Schroeder]

I did research into this subject for my masters degree recently and it came down a multitude of factors. The big one is cost. Many game engines, good ones, cost a lot to license for commercial use. UDK (Unreal Dev Kit) isn't bad if you don't mind paying royalties to Epic every time you sell a product, to a tune of 25% after your first $50,000 in income. If you don't want to pay royalties, you can get a royalty free license for about 250k to 2 million depending on the access to the SDK of the engine as well as what Epic decides your broad reaching goals are. So, for most arch viz, you'd be in the 250k range for a license. Unity isn't a bad engine for it's $1,500 cost but it has a bit of limitations that can take your pipeline down.

 

The second is that arch viz and scene optimization does not go together. Remember, this is real time. You have to be very aware of not only your poly count, but how the model is made. Quads are best, anything else is inefficient to render in real time. All of your textures MUST be a power of 2 (256 ,512 ,1024, 2048 etc) as that's the most efficient texture size to process in real time. Those textures also must be small enough to load and read in real time, yet retain quality. Using Jpegs is an absolute no-no as those are the worst in real time processing. You must understand UVW unwrapping and light mapping channels. You must understand how to build to the grid and build in a modular way. They way you used to build isn't going to cut it as you'll throw your poly/vert count through the roof and lose frame rates.

 

There's some video from my research when I used the UDK engine. It's not bad, but it takes more of a steep learning curve than I think most companies are willing to take on.

[JonRashid]

There are a number of alternatives to udk including quite a number of open source options. As to the UDK licence, surely each project you did would be a seperate executable and thus a seperate product. This would mean that you would only reach their threshold if one job came to over $50,000. But as I said there are alternatives that would only involve you having an initial programming cost of creating an efficient pipeline. Agreed Unity does have limitations but the internet streaming capabilities are quite attractive. As to texture limitations, they are primarily UDK's not all engines. The power of 2 is how a lot of producers already work and most degree students should be doing things like thhat as a matter of course. In a good engine using a fairly standard graphics card 2-3 million polygons and 1-2GB of texture map can be handled quite comfortably on something like an HP200. (standard spec for schools in UK). Agreed optimisation helps, I think any good modeller would agree but as to the suitability of existing models. I have put a number of my old architectural models into our engine at work with the only modification being to any procedural mapping. If you have a look at the latest Open Scene Graph exporter for Studio Max you will see the range of mapping types, lighting, animation(including biped and vertex animation) that you can export directly into an open source engine. The vray users may have to learn some mapping techniques from the games industry but most of them have already been through many incarnations of their work practices. The games industry has shown that excellent results can be obtained with realtime but there are many other industries producing good results. I would love to see what some of the creatives I see on this site could produce within a realtime environment. I am sure the results would be stunning once they became accustomed to it.

[JonRashid]

http://mycryengine.com/ this might up the game a bit too

[Scott Schroeder]

http://mycryengine.com/ this might up the game a bit too

 

If you think UDK is expensive, CryEngine 3 is well, well, well above that. The UDK is 50k globally, not per product. If that were the case, small developers will pull the game from the shelf after the first $49,999 were made. If you sell 10 products at 5k a piece, you are at your limit and the next product you sell from there on out, 25% will go to Epic.

 

This has become pretty much the bible on creating environments for real time: http://www.thiagoklafke.com/modularenvironments.html There really is no better way to create environments that using the modular technique.

 

Scenes from the above tutorial: http://www.thiagoklafke.com/zestfoundation.html

[JonRashid]

There are still some very good open source options. Plus youhave to take into account development costs for your game if that's what your developing. As far as creating content and bibles is concerned for your average architectural development we are not exactly pushing the boundaries of capacity. I can load most of my models onto an iphone app quite comfortably. All I am trying to say is the realtime solution is becoming more and more readily available. Both hardware and software are converging to make this an eminently affordable solution. Perhaps not now for those with no experience of the development process required but times are changing. There are already augmented reality apps of historical sites that run quite complex 3D building environments on phones and notepads. We may be the benificiaries of others developments but it is inevitable that these developments will trickle down through all layers and become both available and affordable. If you look at what the cryengine or UDK are designed to achieve it is way above most architectural requirements and thus is the hammer and nut approach. Just look at Maya - once only used by developers, then the universities obtained for free, now available in schools. Who would have thought 11yr olds would be working with software with the capability of Autocad or Max 10yrs ago. My point being realtime is on it's way and I am sure capable artists will take advantage of this.

[Scott Schroeder]

Oh, I 100% agree with you about real time being right on the doorsteps of arch viz. It just makes total sense to embrace this technology. I mean, just for the video I did in UDK. It was about 10 minutes to calculate the GI maps for the scene inside UDK and the rest was all rendered in real time. Any camera, any move, any location can now be rendered instantly. Why would anyone want to wait hours upon hours for renders anymore?

 

And now, with the CryEngine 3 getting into the mix, the quality is now on par with many true renders.

[JonRashid]

http://www.udk.com/showcase-takenaka

 

http://www.udk.com/news-beta-aug2011.html

 

A couple of examples of people who are adapting to realtime

[Devin Johnston]

Real time hasn't taken hold yet because the quality and ease of use just isn't there yet and neither is the demand. I think the last reason is more important than the other two because clients are usually the driving force behind these trends. Sure it's nice to have a real time model in the office but unless it's used as a tool to inform the client or the design process all it's going to be is a novelty. Clients will have to see the value in it and want it for a specific purpose before they embrace it and right now real time is pretty far behind the ball in terms of usefulness for most people.

[MBetke]

I have to disagree on some of the things beeing said here.

First you don't need to use modular workflow if you use a game-engine for Viz. You don't even have to stay stricktly within polygone limits like if you do a game. The reason is simple: You don't have to optimize the scene for a wide audience. If your client wants an interactive presentation you have to optimize the scene for his target specs and nothing more. Modular building is used to re-use assets on level or even several levels. If you do an interiour design you have your assets and thats it.

 

In engines like Lumion3D you can use any texturesize and you are not bound to power of two sizes. You can even export high-res meshes with over one million polygones with a few clicks.

 

Using real-time engines for projects needs his own workflow. It's a bit of hybrid between developing for a game or doing a 3d rendering scene. For me with my game-dev background i found it easier to spent some polies here and there on scenes as having to optimize meshes all the time.

 

And quality can be really great in real-time if existing tech is used clever like on this project (not mine):

http://www.ronenbekerman.com/forums/finished-work/1139-loft-interactive-unity-3d.html

 

For me it was an advantage using real-time up to now because I can sell it to clients as something new and innovative if they can have their customers walking around in their new home or even exchange objects on the fly (depending on needed 3d solution).