Anybody is thinking about rendering in real time (game engine)

[akbargherbal]

Hi everybody,

 

This is my first post.

 

I am just wondering if the idea of rendering in real time using game engines such as cryengine is viable.

 

Anybody tried...

I would like to see.

[RyderSK]

Search for Tearte interior video. Argentinian studio. They're the only ones who did it nicely (photoreal). Used Unity, and baked in Vray.

I play with CE3 though, it has it's future ;- ) Lot of people already jived into real-time archviz

[akbargherbal]

Thanks a lot Juraj Talcik, for your valuable input.

 

could you provide me with the Argentinian studio, a quick Google search did not suffice.

[Scott Schroeder]

It is a viable option if you do it right, which for most arch viz people/studios represents a change of direction for their pipeline. You must start to think like a game developer and less like an arch viz developer. Exact measurements don't matter much.

 

Then again, you have licensing cost issues and CryEngine may be free to use but it will cost you to make money from it. Ditto with UDK and to a lesser extent Unity. Unity can be used freely to make money until you make $100,000 US in one year and then you MUST pay for Unity Pro. You cannot mix and match licenses with Unity, so either everyone is on Pro or everyone is on Free.

 

Baking textures from Max is the most inefficient way to use a real time engine and tax your video card. You might as well not even use it. You drastically increase your draw calls, use over sized textures and many other things. Do not fall into that trap. You want to be able to re-use as many textures as you can to increase your frame rate. Having a unique texture per object isn't very effective and takes time to create every texture and apply it. There is a reason why most of the walk through done in this fashion play at 10-15 FPS rather than the real target of at least 30 and probably better at 60 FPS.

 

This guy used NO diffuse textures:

 

This person used one 256X512 sheet for everything.

 

And, of course, my own personal work with UDK

 

You really want to do more with less when it comes to a real time pipeline, and this is even more important if you are wanting to use a game engine. Engines such as Luminon are a little more forgiving in terms of direct exporting of massive meshes and textures.

 

As I said, it is a viable option. However, it does involve a substantial shift in your method used to create the environment and a lot of extra training is needed. So, in the real world, it is not as good of an option due to the sheer extra time requirements.

 

What is nice, using my video as an example, it took about 45 minutes to bake the full quality light inside UDK. Once that bake is done, you never need to do it again (unless you move objects, you can change textures though) It took the time of the animation to render out the frames. So a 3 minute animation takes 3 minutes to render after the light bake. I can now add any camera in any location and render out new frames in as long as the sequence is.

 

Polycount is a fantastic wealth of information for real time environment creation:

http://wiki.polycount.com/CategoryEnvironmentModularity

 

You'll find more well rounded information there from people inside the game industry who use these engines on a daily basis, and not just toy with them like I do.

 

If you are thinking about this route, take your time to research it and do it right.

[RyderSK]

I have to disagree. You give advice like you would to starting indie game developers. Archviz market demands absolutely different quality, different approach, different result.

 

You don't have to think like game developer at all, you have to KEEP thinking like architectural visualiser and come up with ingenious solutions how to get highest possible quality in quite constricted real-time environment. That's the real creativity here.

 

Just compare your videos (which in my opinion have zero relevance to this thread or real-time archviz at all, there is nothing to take away) with this:

 

 

Do you think clients are interested in "Mass effect" type visuals for their architectural projects ?

 

I won't comment on your UDK example, that is highly amateur and ugly.

 

Game industry approach=/= archviz realtime approach

 

With kind regards, someone, who was asked to work for DICE(Battlefield 3, Mirror's edge, arguably two games with the best visuals in industry, where 90perc. of stuff was baked in outside of engine's capabilities, which does feature quite advanced GI, unlike UDK) as lightning artist,

 

Juraj

Edited June 27, 2012 by RyderSK

[Scott Schroeder]

So why can't this industry, arch viz, present a decent example for real time? What is this industry doing wrong that obviously the games industry has gotten right? All of the best examples for real time arch viz are very small projects in scope. They are proofs of basic concept at best.

 

Your example you show, while done well, is just one small studio apartment. How does it hold up to a home? How about exterior? Skyscraper? How about anything less than spartan furnishings? What about landscaping? What if you want to move an object, does it leave a burned shadow?

 

Is my video the best? Heck no times infinity plus one! It has a whole host of things wrong with it, I have no issue with having that pointed out. However, it was a headfirst R&D dive into the world of real time mixed with game environment modeling techniques. It was merely as an example of what is out there at this current moment.

 

Games have one thing that will almost always be an anchor to real time arch viz and that is resale. EA/DICE raked in billions with Battlefield 3 so they could afford a $200 million plus development cycle stretched out over a few years with a team of hundreds. Can't really get those kind of numbers in this industry. The client you are making this for can't go resell your tour to make money. It's a one time use item, and that is a severely limiting factor. The other big this is refinement cycles. In architecture, we are all too often looking for push button solutions that will take a overly complicated model and spit out pretty real time images.

 

That simply will never happen. The Dallas Cowboys stadium that was done in UDK is a prime example. Watch that video and it's clearly an example of why not to use real time. It was awful, lagged like crazy, and was overly repetitive in nature. Why was it so? They just dumped a 38 million polygon model in there and expected it to work. That's thinking a lot like architects and less like game developers if you ask me.

 

There are so many approaches to this. I never expect anyone to fully agree on one process, set of ideas, or engine to use. We can agree Juraj, (hopefully) that real time is a severely untapped tool for architecture though.

 

Here's a few more traditional architecture based projects from UDK:

http://www.unrealengine.com/insiderblog/architectural_design_with_the_unreal_engine

Tripod 3D has some pretty interesting stuff on there. They seem to be closer to using real time than most places.

 

So for now, in the terms of traditional game engines, the top three that I've seen in terms of popularity and quality are Unity, UDK, CryEngine 3. Frostbite 2 is off market as far as I know. If it is available for licensing, it is beyond the price range of any viz studio.

 

A wonderful 5 part series on lighting in Battlefield 3,

 

Notice that Mirrors Edge used Unreal 3 Engine by the way, which is what UDK is using.

http://www.shacknews.com/article/52569/mirrors-edge-built-with-unreal

[Dimitris Tolios]

I've actually used and liked LumenRT for simple projects. Yes, it's Sketchup based atm, but does go a long way imho into providing a GI look through real time D3D acceleration, and the "live cube" concept is great for sharing simple pre-defined and at-any-time interactive walkthroughs with your clients. It is dirt cheap too (for any serious professional).

 

It pretty much repackages your 3D model and translates your shaders into D3D, precalculating GI with your CPU for the whole scene (set time of the day). Then you can create an animated sequence out of it, where it simply animates between your saved SU scenes (which you can arange into a pretty decent walk-through sequence), as your GPU renders the output in real time. At any given time you can use your mouse and/or arrow keys to walk and look around the building.

 

The results some times are not that sophisticated, but so isn't the overall interface: it uses simple word triggers to give some extended attributes (reflection / bump map / water etc) to existing SU materials, and then you just choose time of the day and 1-2 more options...it does not like huge or really complex models tho...i would love to do some fly-by videos with it, but urban scale site models crash it.

 

Download the sample videos, the results are pretty nice for a SU only generated content.

 

Lumion 3D does pretty much the same job, but it does get a bit more sophisticated.

 

Both can provide results that anywhere match to blow away games - given you don't focus on env effects and explosions.

Edited June 27, 2012 by dtolios

[Mario De Achadinha]

Hi Juraj

How big is the final file that you open with this realtime engine?

[neil poppleton]

Have a look at Promenadd 360 software : http://forums.cgarchitect.com/72724-promenadd-3dsmax-panorama-hd-360-plugin.html#post371920

[Mario De Achadinha]

Great Link neil!!!! Just one question is it for iray or does it support vray too?

[neil poppleton]

Hi Mario

 

We use with vray and use the power of 3dsMax network rendering to produce. It is a great simple but powerful tool.