E-on-Lumenrt render for Sketchup

[Ecuadorian]

By examining the available info, this is what I either know or especulate so far:

LightUp: No need to export
Lumion: You have to export via Collada. Sometimes materials get all messed up on re-import in Lumion.
LumenRT: Seems to have its own SU exporter which then fires up LumenRT on a separate window.

LightUp: Needs pre-processing time. Renders 720p video @ 3 fps with an average GPU. No speed gains from top GPUs.
Lumion: No pre-processing needed. Renders 720p video @ 0.3 fps with a nice GPU (GTX 570)
LumenRT: Needs pre-processing time. Renders 720p video @ 30 fps. Hardware to achieve this: Unknown.

LightUp: Can use several cubemaps, called "irrcaches", for reflections.
Lumion: Can use only one cubemap for reflections. You can choose where to generate it if you want, or it can be generated from the camera's point of view.
LumenRT: Seems to be using either real-time raytracing or flat-mirror maps. Either way, it looks much better.

LightUp: No actual bounced light calculations. Just pre-calculated AO with color bleeding.
Lumion: No actual bounced light calculations. Just SSAO with no color bleeding.
LumenRT: Seems to be pre-calculating bounced light with color bleeding.

LightUp: Can show direct light from emitters and point lights.
Lumion: Can not show direct light from emitters and point lights.
LumenRT: So far, no direct light from either emitters or point lights has been shown in the previews.

LightUp: Shadows are saved in the mesh.
Lumion: Uses shadow maps.
LumenRT: Seems to be using shadow maps, too.. or raytracing them in real time.

LightUp: Can export a model you can freely explore in LightUp player format.
Lumion: Can't export a model in any way.
LumenRT: Can save an executable file you can give to a client. No need to install any player.

Can't compare to Twinmotion as I haven't used it.

[Gaieus]

Yes, this is pretty much that I can gather, too (or even less of the technical stuff at which you are much better of course).

[dburdick]

Hi everyone,

Dave Burdick from e-on Software here. We're pretty pumped up about LumenRT as it will be released soon. I thought I would jump in here and add my 2 cents to this very excellent post:

@ecuadorian said:

By examining the available info, this is what I either know or especulate so far:

LightUp: No need to export
Lumion: You have to export via Collada. Sometimes materials get all messed up on re-import in Lumion.
LumenRT: Seems to have its own SU exporter which then fires up LumenRT on a separate window.

Yes, LumenRT has a built-in custom Sketchup exporter which automatically exports the Sketchup model and then fires up the LumenRT renderer.

@ecuadorian said:

LightUp: Needs pre-processing time. Renders 720p video @ 3 fps with an average GPU. No speed gains from top GPUs.
Lumion: No pre-processing needed. Renders 720p video @ 0.3 fps with a nice GPU (GTX 570)
LumenRT: Needs pre-processing time. Renders 720p video @ 30 fps. Hardware to achieve this: Unknown.

LumenRT runs on any modern PC or Mac but rendering speed strongly depends on hardware. On my 4 core Intel 2.4ghz machine for example, a Review mode render (good enough for most cases) takes 15 minutes to an hour to render out a model depending on the size and complexity.

@ecuadorian said:

LightUp: Can use several cubemaps, called "irrcaches", for reflections.
Lumion: Can use only one cubemap for reflections. You can choose where to generate it if you want, or it can be generated from the camera's point of view.
LumenRT: Seems to be using either real-time raytracing or flat-mirror maps. Either way, it looks much better.

Yes, we compute real-time reflections.

@ecuadorian said:

LightUp: No actual bounced light calculations. Just pre-calculated AO with color bleeding.
Lumion: No actual bounced light calculations. Just SSAO with no color bleeding.
LumenRT: Seems to be pre-calculating bounced light with color bleeding.

Yes, that's correct.

@ecuadorian said:

LightUp: Can show direct light from emitters and point lights.
Lumion: Can not show direct light from emitters and point lights.
LumenRT: So far, no direct light from either emitters or point lights has been shown in the previews.

Yes, only the sun - no other light sources.

@ecuadorian said:

LightUp: Shadows are saved in the mesh.
Lumion: Uses shadow maps.
LumenRT: Seems to be using shadow maps, too.. or raytracing them in real time.

They are real-time shadow maps for increased accuracy.

@ecuadorian said:

LightUp: Can export a model you can freely explore in LightUp player format.
Lumion: Can't export a model in any way.
LumenRT: Can save an executable file you can give to a client. No need to install any player.

Yes

[Ecuadorian]

Thanks for the very informative reply, Dave.

A friend in another forum is asking whether this will be GPU-based. Based on your response, seems like pre-calculations are currently computed only on the CPU, while navigation will use mostly the GPU via OpenGL or DirectX, with the CPU handling real-time reflections and shadow maps. Am I correct?

I'm used to rather long pre-calculation times thanks to LightUp. Usually 35-40 minutes for a house like this one, using a 10cm mesh. However, Adam Billiard, its creator, has stated in LightUp's forum that he plans to move the pre-calcuations from the CPU to the GPU, using it as a "compute resource", "like Octane" (Maybe CUDA or OpenCL ). When this is implemented, there will be "no waiting", according to him.

Do you plan to move LumenRT's pre-calculations to the GPU to reduce waiting times, as well?

Also, I guess a "LiveCube" (I'd prefer "pre-lit model" as "LiveCube" sounds like a simple panoramic still image) will reside in the GPU RAM, so the more complex the model and the finer the illumination mesh, the more GPU memory you'll need to be able to display it, right?

The need to pre-calculate and store this information will also mean there's a limit to the size and detail you can attain. If you want a bigger size, you'll need to lower the detail of the illumination solution. I also see no mention about a dynamic sun/sky/cloud system, object animation capabilities, or even an object library. These limitations would mean LumenRT will be directly competing with LightUp ($150) instead of Lumion (750€), so I hope you guys price it sensibly.

[dburdick]

Hi there.

@ecuadorian said:

Thanks for the very informative reply, Dave.

A friend in another forum is asking whether this will be GPU-based. Based on your response, seems like pre-calculations are currently computed only on the CPU, while navigation will use mostly the GPU via OpenGL or DirectX, with the CPU handling real-time reflections and shadow maps. Am I correct?

Yes, this is correct the CPU computes what we call the LumenRT "Live Cube" and the the GPU displays it and allows you navigagte around the Live Cube.

@ecuadorian said:

I'm used to rather long pre-calculation times thanks to LightUp. Usually 35-40 minutes for a house like this one, using a 10cm mesh. However, Adam Billiard, its creator, has stated in LightUp's forum that he plans to move the pre-calcuations from the CPU to the GPU, using it as a "compute resource", "like Octane" (Maybe CUDA or OpenCL ). When this is implemented, there will be "no waiting", according to him. Do you plan to move LumenRT's pre-calculations to the GPU to reduce waiting times, as well?

It's possible. As GPU's and GPU software like CUDA/OpenCL become more powerful and more mainstream, the heavy graphics computing tasks will be offloaded to the GPU. I don't think we are quite there yet unless you own a very high-powered GPU cluster like an NVIDIA Tesla card setup.

@ecuadorian said:

Also, I guess a "LiveCube" (I'd prefer "pre-lit model" as "LiveCube" sounds like a simple panoramic still image) will reside in the GPU RAM, so the more complex the model and the finer the illumination mesh, the more GPU memory you'll need to be able to display it, right?

Yes, this is true. Most midrange graphics cards such as the the NVIDIA GeForce 460 can handle faily large models with 512 mb of graphics memory or more.

@ecuadorian said:

The need to pre-calculate and store this information will also mean there's a limit to the size and detail you can attain. If you want a bigger size, you'll need to lower the detail of the illumination solution.

Not really, it will just take a longer time to render (e.g. pre-compute). If you have a midrange GPU card it will handle most typical models just fine. If you include all kinds of fine modeling details with modeling polycounts above 500K or so, you may see some degradation in performance on the GPU side.

@ecuadorian said:

I also see no mention about a dynamic sun/sky/cloud system, object animation capabilities, or even an object library. These limitations would mean LumenRT will be directly competing with LightUp ($150) instead of Lumion (750€), so I hope you guys price it sensibly.

The first release of LumenRT will not include dynamic skies. Camera animation is done by exporting the Sketchup scene animation into LumenRT but there is no separate object animation system.

[Ecuadorian]

Sounds fairly similar to LightUp so far, but with bounced light, better reflections, no artificial lights, and external instead of internal. After exporting, do you adjust settings/add objects or is everything set up before exporting? If you do alter some settings after export, how will changes to the SketchUp model be handled?

I have to say it's fairly common for me to go beyond 600k edges/300k faces in models I've animated with LightUp, like this one. However, since LightUp uses a finite-element method (aka a grid) to solve illumination, I can't use it to create a walk-through of a 1Km x 1Km housing development, as I would need to use too wide a grid to show any light detail in close-ups. I only use LightUp for compact models. Seems like LumenRT, which seems to also be using a finite-element method, will be confined to similar compact projects (houses, restaurants, chapels), as so far the examples have been fairly small spaces. No house developments, no large shopping malls, no stadiums have been shown. I guess one would still need Lumion for larger projects. As it uses a screen-space effect, it can handle anything you throw at it and it can still look good on close examination, no matter the real-world dimensions of the project or if it has several million edges.

Here's a sample of my walk-through work with LightUp, Lumion and Twilight. I use different apps for different needs, because so far I haven't found one that can handle all kinds of projects. My guess is that LumenRT will be handling the individual houses part, basically becoming an alternative to LightUp.

[dburdick]

Hi Ecuadorian. Very nice animation you made there.

@ecuadorian said:

Sounds fairly similar to LightUp so far, but with bounced light, better reflections, no artificial lights, and external instead of internal. After exporting, do you adjust settings/add objects or is everything set up before exporting? If you do alter some settings after export, how will changes to the SketchUp model be handled?

There are no adjustments after export into LumenRT. The process is entirely seamless. You just indicate the rendering quality you want ranging from draft (quick preview render taking under 1 minute), Review (~30 minutes or less), Final (~ 1 hour or less), Superior (~ 2 hours or more) and the type of atmosphere you want and then LumenRT takes care of the rest. There is no ability to add objects or new camera animation paths post export into LumenRT. If you need to change something on the model, you do it in Sketchup and then re-export into LumenRT.

@ecuadorian said:

I have to say it's fairly common for me to go beyond 600k edges/300k faces in models I've animated with LightUp, like this one. However, since LightUp uses a finite-element method (aka a grid) to solve illumination, I can't use it to create a walk-through of a 1Km x 1Km housing development, as I would need to use too wide a grid to show any light detail in close-ups. I only use LightUp for compact models. Seems like LumenRT, which seems to also be using a finite-element method, will be confined to similar compact projects (houses, restaurants, chapels), as so far the examples have been fairly small spaces. No house developments, no large shopping malls, no stadiums have been shown. I guess one would still need Lumion for larger projects. As it uses a screen-space effect, it can handle anything you throw at it and it can still look good on close examination, no matter the real-world dimensions of the project or if it has several million edges.

A couple of things here:

1. LumenRT does not use a grid/finite element method to calculate lighting. It can handle models of any physical size such as a 1km x 1km housing project. However, it will take much longer to render – perhaps several hours. The display of the final output will of course be totally interactive.

2. LumenRT was designed to be used as a collaborative tool for sharing design concepts, ideas, and models with your clients in a more immersive, interactive fashion. In other words, your clients will now have the ability to explore and navigate the models you create themselves, with very high quality lighting and fidelity. It’s not meant to replace tools like LightUp or Lumion or any other rendering tool that produces canned or fixed output. It’s really a complimentary tool or another “arrow in the quiver” to provide a way for you to engage with your clients in a more immersive, interactive fashion – to let your clients “play with” or “explore” the design.

3. The only thing you need to be mindful of when using LumenRT is that the interactive GPU performance is of course dependent on the capabilities of your client’s graphics card. Our tests have shown that for models with 500,000 polys or less, there is no performance issue at all provided you are using a reasonably modern graphics card. For your model shown in the link this one, this would probably exceed 500,000 polys and therefore experience some degradation in terms of FPS performance.

[Ecuadorian]

The reason I believe LumenRT and LightUp will be close competitors is that LightUp (and soon Lumion) can also export models that you can send to a client to freely explore. The LightUp player is free and only a 367KB download. I can navigate the model you saw in the video @1080p in LightUp player with no performance hiccup on a Nvidia GTX 570. Models with a lot less detail can be explored on mainstream GPUs.

However, in my experience, you never know what kind of computer the client has, if he's tech savvy enough to open an exe, bypass Windows' UAC and antivirus warnings, and navigate a model you send him without getting lost or getting stuck in a corner. Maybe he doesn't even have computer at all... some people only use smartphones nowadays and have forgotten computers. So the only time you end up using freely explorable models are when you take your laptop and navigate the model yourself in front of the client.

That's why I prefer videos and stills even though LightUp (and soon Lumion according to its creators) also offer navigatable models.

[dburdick]

Hi Everyone,

Here are some links to a few new LumenRT renders. I really like the softness and richness of the light in these:

http://www.lumenrt.com/images/showcase/CafeRosso01.jpg

http://www.lumenrt.com/images/showcase/CafeRosso04.jpg

http://www.lumenrt.com/images/showcase/FelixHotel_03.jpg

http://www.lumenrt.com/images/showcase/FelixHotel_05.jpg

[AdamB]

@ecuadorian said:

The reason I believe LumenRT and LightUp will be close competitors is that LightUp (and soon Lumion) can also export models that you can send to a client to freely explore. The LightUp player is free and only a 367KB download. I can navigate the model you saw in the video @1080p in LightUp player with no performance hiccup on a Nvidia GTX 570. Models with a lot less detail can be explored on mainstream GPUs.

However, in my experience, you never know what kind of computer the client has, if he's tech savvy enough to open an exe, bypass Windows' UAC and antivirus warnings, and navigate a model you send him without getting lost or getting stuck in a corner. Maybe he doesn't even have computer at all... some people only use smartphones nowadays and have forgotten computers. So the only time you end up using freely explorable models are when you take your laptop and navigate the model yourself in front of the client.

That's why I prefer videos and stills even though LightUp (and soon Lumion according to its creators) also offer navigatable models.

If you're concerned about your clients hardware, you can simply export a panorama from LightUp with a single button click:

[flash=512,512:2o630022]http://www.light-up.co.uk/resources/cubicvr/panoviewer.swf?filePath=http://www.light-up.co.uk/resources/cubicvr/cube1/&prefix=cofla[/flash:2o630022]

Also ran the Cafe Bar model through LightUp v2.1 after adding a bunch of IES lights sources + some area light sources around the bar area (LumenRT just does a single parallel light ie The Sun, right?)

Lighting time was 5m40s. Images straight out of LightUp with no touching up.

and and quick daytime one too:

[miikka1978]

Now there seems to be more information about Lumenrt in their webpage.
http://www.lumenrt.com/

[dbalex]

yes but it went away as fast as it came online :s

[d12dozr]

No, just moved to a different page...its available for purchase now! http://www.lumenrt.com/purchase/
Special intro price of $195.00...SRP $295

[Ecuadorian]

So, who'll be the first to bite this time?
I have found no free trial...

[Ecuadorian]

First observation: Reflections seem to be hybrid. For closer objects it seems to be using either real-time raytracing or a real-time flat mirror map. Far away objects seem to be rendered using a standard reflection map, either cubical or spherical.

[Panga]

It seems that QuadroFX cards are not supported ! I tried to launch one of the boxes available for download on their website, and it's not workink with my Quadro FX 1800...

[raxport]

@ecuadorian said:

So, who'll be the first to bite this time?
I have found no free trial...

Hello there sketchers. I'm new here on this forum.

Well i did bite

I'm currently trying out my new gadget... it will be interesting.

Sig

[thecravatman]

@panga said:

It seems that QuadroFX cards are not supported ! I tried to launch one of the boxes available for download on their website, and it's not workink with my Quadro FX 1800...

I have no problem and use a Quadro FX 1700. I do have 8meg of ram. It works beautifully for me. All the models I have rendered are acceptable most really good. Beware high quality can take an age. Plants need to be low polygon.
I will definately use this as part of my SU Vue Max pipeline. I would like a range of materials that can be applied in SU, work really well in LumenRT and then translate to Vue materials when imported into Vue so you really get continuity.
I hope that is of interest.

[Panga]

Thanks for your feedback thecravatman ! Maybe I need to update my nividia drivers...I'll try again tonight.

Cheers.

[asulaiti]

Hello to all

I would like to thank you first on this subject with all the information it .. I am interested in the LumenRT program, but I don't know more about it, or about all rendering plugin for SketchUp.

I use Cinema 4D, and plan to learn a second program, to improve some projects, and the Google SketchUp is one of my choices.

My question is focused on how we comparison between LumenRT and LightUp, Or they are completely different?

If they were in the same field .. What the best choice for purchase?

Any advices I will be thankful.