First thoughts about Lavina, and wishlist

[chroma]

Hi! I finally did some testing with Lavina, and my thoughts are:


- vrscene seems like a culprit. Somehow there should be an "update" button from 3ds max, so that the affected parts of the scene updates, while the rest is maintained as it is. Exporting a .vrscene and then importing in Lavina for a small project change seems backforward.

- Look at the attached images grass1 and grass2. Not a beauty scene, nontheless 2 is from max and 1 is the same scene in Lavina, using the same hdri for lighting. Not 100% comparable, but the Forest Pro-grass in Lavina seems too dark. I haven't researched this further, but found it strange.

- Basically a lot of the functionality that Twinmotion has, but in particulary:
1) dynamic atmosphere, cloud and sun system, with different cloud types, density and height
2) animated 3d scanned people with poses
3) the mentioned update link
4) export/build to hmd's like oculus quest

More thoughts:
I see this as an extended render program. Like mentioned in this forum, Lavina would probably be too heavy if all vray functions would be implemented. But render passes, nice intuitive animation functionality.
In my dreams, I'm building the scenes in 3ds max, export the project to Lavina for setting up renderings with quick feedback on dof, camera angles, lens effects, and so on, and hits the update button if something in my host scene has changed.

Or even better:
I am not sure why you don't want this to be a competitor to vray. I think you should implement Lavina into vray, and then just make it possible to turn on features if their workstation allows to.
Then you can, in the very same 3dsmax vray setup, just by changing the render mode, take the scene from the "building" process to the render process - instead of forcing people to export to a static scene that must be imported. That is not the way it should be done in 2020. Vray will eventually have to get to this Lavina stage to survive anyway.
Just my two cents.

As Twinmotion is also coming with realtime raytracing this year, they seem like a bigger competitor for you in the archviz field than vray itself. And that's a community you'd like to keep, I suppose?


- What are the developers and strategists at chaos' thoughts about the potential for Lavina? Where are you aiming it, and what does the roadmap look like?
Would you rather leave it just as a viewer, then I believe you have lost the game. It needs some functionality that eventually (very quick I suppose) will overlap vray.
And since Vray is already a big name, it seems for me like a better take on this to implement it into Vray.


But otherwise it's easy to forget that it's realtime raytracing. Good work.

[Phillip Miller]

Thanks for the feedback Croma – that is what this Beta is for!

>> but the Forest Pro-grass in Lavina seems too dark
We are looking at the Forest Pack grass issue, but from the looks of it, it seems the textures are not loaded. This is most likely due to not supported texture, e.g. the Forest Color. Do you mind showing the material network or sharing an example scene?

>>…update button if something in my host scene has changed.
We are looking at ways to make this more streamlined – stay tuned.

>> export/build to hmd's like oculus quest
It’s on our list to provide an API for direct support of devices.

>> animated 3d scanned people with poses
This isn’t on our list for this year.

>>What are the developers and strategists at chaos' thoughts about the potential for Lavina?

Project Lavina is its own application so that it can come to market as quickly as possible while also being liberated from the many overheads of being within a creation tool. It is not trying to duplicate what game-engine solutions provide, but instead is doing what they are not capable of doing – making massive scenes real-time, and with no conversion or optimization work. It also allows us to develop a common environment that is immediately accessible from all V-Ray products – from SketchUp to Houdini.

If you’re using large scenes, you probably already have noticed that they are loading much more quickly than they did where they were created – and then they are probably far more interactive than they were in the creation viewport as well. This a benefit of pure ray tracing – where performance is fairly constant for a given resolution, regardless of how much geometry is being handled. Project Lavina then does its own magic in handling tons of light sources while staying fast. This is the opposite of a game engine, which gets progressively slower with more geometry or lights. With Lavina, you can smoothly explore scenes that are painfully slow in a creation tool and won’t even load in game engine.

So that’s where we start – super fast ray tracing, with original geometry – where the concept of “optimizing” can be forgotten. We are now adding the “non-rendering” functionality for being productive in this new world. It is also out intention for you to be able to take your Lavina decisions farther – either to V-Ray or back to the creation tool – and we’d like everyone’s priorities on what’s most important here.

>> I think you should implement Lavina into vray…
In time, you will probably be able to access the Lavina renderer from our V-Ray integrations, but that isn’t the focus of this Beta. We are currently making Project Lavina a valuable companion to your creation tool, so that’s what we would like everyone’s feedback on. We otherwise do see Project Lavina as an extension of the V-Ray ecosystem, as it does leverage your V-Ray setup work – so if you are using V-Ray, you can always take your work to real-time with minimal effort.

- Phil

[chroma]

Thank you so much for your response! Yep, Forest Color. Spot on. Good that you're on the case!

Do you see Lavina as a "show the customer realtime to take decisions in the meeting room/streaming the result over screen" type of application, or do you also see this as a renderer where you by time can export animations and loads of images with less time (given you have good gpus). The question might seem stupid, but I don't find the answer to be quite obvious. Now, I'm on a 1080ti only, so the render times in Lavina won't count, I know, but anyway it was just as slow as one of our nodes. I'm getting a 2080ti rtx soon and looking forward to test Lavina with that one.

Also in this time of threadrippers, I believe I read something about Lavina hybrid support not gonna happen due to the platforms, but look at it this way: expanding our render farm with a few more 48 core threadrippers would almost be the same price as trying to leverage the big scenes into the gpu memory (two, three four titans for our biggest scenes). I mean, things are happening in the CPU world as well with more and more threads. IF the 2080t and upwards cards keep their price in the market, it would be pretty expensive running a 128 gig equivalent on GPU.

And also another question for you. How is the memory allocation/handling? Would it swap somehow, or am I litterally in deep waters if I've got a huge scene with 50-60 high detailed houses with all vegetation, cars, people, huge textures, you name it?
I also did notice another thing. Is Lavina doing the instancing correct? My vrscene test scene got sooo big. It might be just me not having control over all used assets, but it seemed overly huge, resulting in the system to crash.

I couldn't help noticing that my HDRI scene light didn't export/import. That is a must, to recreate what you've been doing in Vray. Also it needs to bring in the camera with all its parameters.

What about dynamic sun/cloud system, wouldn't that be really great game changer, instead of having to load a hdri? A scene viewer like this should be able to use both it's imported light, but also a flexible dynamic cloud system with real sun positions (which also can be detached from reality and controlled custom). As raytracing is all about light source...

Uhm, and doing quick changes in the VrayMaterials back and forth from the scene in max (or other). Or just doing that udate quick in the vray native program for then do update, but that seems like an unnecessary step?
That was a bunch more questions...

Thanks for the opportunity to beta test this product, it's really interesting.

[Phillip Miller]

>>Do you see Lavina as a "show the customer realtime to take decisions in the meeting room,,,,or do you also see this as a renderer where you by time can export animations and loads of images with less time.

Yes and yes. Lavina's primary goal is to be the best, and easiest, way for you to explore your design. In doing that right, it also becomes a great way to communicate with others and present. And because we can't help but do great rendering, the actual output is good enough for a lot of situations. But in most cases, customers are telling us what they want to do with Lavina rather than us telling them how to use. And that's why we're doing this Beta

In terms of processor support, the amount of performance to be gained from supporting the CPU would be minimal in proportion to RTX - especially when such a large task would keep us from simply making our GPU code that much faster. Unlike V-Ray GPU, Lavina does not view render farms (which are nearly only CPU) to be a target. Lavina is about interactive experiences - which means you are working on a dedicated graphics workstation of some sort. We applaud AMD for making such huge strides on its Threadripper line, and V-Ray GPU is right there to lap it all up. But we are fine with Lavina being GPU-only, and keeping that focus allows us to evolve it more quickly. It also makes it much easier for people to adopt - because a new CPU nearly always means a new motherboard/machine, whereas a new GPU is a 5 minute swap or an eGPU add-on to most Windows 10 PCs or Laptops.

>>How is the memory allocation/handling?
At this time, your entire scene (geometry and texture) must fit into GPU memory. There are some other discussions here for details on how to manage that and what we are doing to improve it.

>>Is Lavina doing the instancing correct?
Yes.

>>my HDRI scene light didn't export/import.
I believe this is on our list to fix soon.

>>What about dynamic sun/cloud system
It's on our list to support V-Ray Sun & Sky.

>>back and forth from the scene in max
We have several ideas on possible workflow improvements for making this easier. Please look for a post on this - asking for opinions - in the next week or so.