Materials

[Naiman Broome]

Hi I would like to know if are previewed also introduction of materials with the meshes , to my opinion meshes without a proper bit alpha sorting and materials to give at least a minimum selective shine or bump are very much limited in their look and proper use ... 

 

What are the specs about this subject?

[Drongle McMahon]

The developers have indicated that materials will not receive further attention before release, but their comments have indicated that it has a good chance of being given a high priority after release.

[Nyx Linden]

You've correctly identified one of the projects we'd like to get to post-release!

Such a project would include VWR-6713 as you indicated in your other thread.

Thanks for the interest in making our content creation tools better!

[Indigo Mertel]

Hi Nyx,

what are the chances that Linden Lab may adopt a technology like Allegorithmic Substance in the future?

Parametric materials with a very small size footprint. A product like Substance seems to be a perfect match for a virtual world like Second Life.

[DanielRavenNest Noe]

Not necessarily.  Procedurally generated scenes  are  a powerful tool, but they have several drawbacks.  One is that they require more processing at the user's machine to build the material or geometry out of a procedural formula.  Another is applying a static texture is simple to understand.  In a world with user content creation, using complex node-based generators like Allegorithmic's product may be too hard for most people.

Here are two scenes I created using two different editors that use "procedural" methods, to show what can be done:

http://lh6.googleusercontent.com/-KUtb9rlDl2s/TeEBUkBOPcI/AAAAAAAACiA/tUQID-c_9xY/s1440/3%252520hills.jpg

The first only has about 6 types of object in the whole scene: fractally generated terrain shape, fractally generated terrain texture, a tree, and three kinds of grass. The grass is randomly scattered by "painting", so each tuft only needs a location, but all the tufts are copies of one of the three models. The trees I placed by hand to make a nice photo.

http://lh4.googleusercontent.com/-jqfcK4qWpEM/TapoP3mOhSI/AAAAAAAACfM/CCtJaSbK1t4/s1440/TerrainExample.JPG

The second has 3 objects: a tree, a bush, and a grass, and the settings on the right allow applying them to the whole of a terrain area like the grass or sand, but with parameters for altitude, slope, and density to control where and how many get used.  The distribution on the map does not exist in saved form at all.  The plants are instanced as needed within your view distance as you move around.

So these examples show you can get cool results without using a lot of items, but the methods are not as easy to understand as applying textures to a prim, and they require a fairly good graphics card to implement (at least for these examples).

[Indigo Mertel]

Thank you for your comments, Daniel. Those images look great.

I am not an expert in this field, so forgive me if I say things that don't make sense. However, thinking about it I am not sure adopting such a technology would be so remote.

As a start, the two texturing methods would have to co-exist for compatibility with the existing content. Thus, builders would have a choice. Also, procedural texturing would be more akin to builders working with 3d tools for mesh, people who have the expertise to make it work. And, judging by videos I have seen on the use of Allegorithmic Substance in 3dMax, the use of it doesn't seem to be out of reach.

For what I understand, another advantage of a product like Substance is that each material carries its own definition with itself and parameters can be edited. This is evident with the free Substance Player, where a user can tweak the material. In SL the player could be integrated into the viewer, just as Windlight is.

When you consider that many builders don't create their own textures, rather they buy them from texture artists (I am one of them), I don't see why this could not work for procedural textures as well. Those with the skill create and sell them, those without buy them with the advantage of having a material that can be tweaked to their liking. Tweaking a material as it is done with the Substance Player is a rather simple process, it's just a matter of working on some sliders to change the parameters.

As for performance, William Burns has a very interesting post on the Andromeda Media Group's blog on Generative Modeling Languages (GML). Quoting William: "with Generative Modeling, we’re talking about a procedural methodology that can just as easily generate more or less detail based on the hardware running it, but do so with the same information regardless". A very interesting approach.

What if Linden Lab developed its own version of the technology, simplified and optimized for its needs? Or, licence the technology? Or, purchase an existing company like it did with Windlight? The advantages would be significative, both in terms of rendering quality and optimization.

[Aeonix Aeon]

As an aside, the concept I was referring to in relation to procedural textures is that if we are generating the textures from the same underlying data, by which we are extrapolating the detail as a mathematical routine, then the lower fidelity takes less time to generate versus the high fidelity.

In a static world, we're choosing a preset level of fidelity and forcing the clients to generate to that level (or preset levels) which I would readily agree would be costly for lower end systems to maintain. However, the idea that I was getting at was that procedurally generated textures can be generated with less as well as more fidelity based on computation time given to the routines up front, in which case we can create an automatic adjustment routine tied to something like FPS wherein if the GPU load is too high and the FPS drops below an acceptable threshhold, then the level of detail for the procedural texture generation is lowered to compensate and balance the FPS versus Fidelity.

In this manner, we create a procedurally generated world (at least dealing with textures) wherein the same data used to represent the high end movie-quality rendering on gaming rigs can be used for lower end systems with less streaming computation. While the fidelity of the output does lower in the process, it is an acceptable trade off in much the same way we wouldn't expect all systems to comfortably handle Ambient Occlusion and Shadows maxed out in quality.

As far as I am aware, systems such as Allegorithmic Substance stream the procedural computation through the GPU in real time based on similar criteria, as well as lower the fidelity of the output based on radius from the camera (because obviously we don't need high definition output for items in the distance). The end result is real-time procedural texture streaming that can scale up and down in fidelity based on the client computer running it.

 

Thank you for the mention, Indigo

[Poenald Palen]

I have not read anything of new materials. I have mentioned it way back on the old xstreetsl forums...wow, that was a while back. I figured they need more shiney control and an ability to use a different texture for bump mapping. I think I also complained about not having specular maps.

If you think about it you can see the ease in just giving more shiney control, another bump texture other than the one you are using for main texture and maybe some kind of more advanced bump map at that. I mean, that all seems pretty easy enough (well, easy compared to making a procedural system or making some kind of harness mapping and allowing like 10 layers with each layer allowing different reactions (specular, harness, diffuse, glow, etc.) and so on. Just adding ONE change to allow a texture that is not the main texture is easy compared to all that. Adding more control to the existing shine is not so hard at all. So....well....I can only assume this is the direction they might go.

I have seen the new shiney in the new lighting and shadow rendering and you can see that the new shiney and light and shadow all sort of are a beginning to a new look in SL and the full potential of windlight being seen/realised. Hey, add to that the DOF camera fuzziness (wow, that stuff is neat. I smiled when I saw this one.

SL is moving toward becoming a leader in mechinima, IMHO. The camera control, the lighting and now features like DOF are all very much stuff that is going to make much more artistic stuff. Any new materials help to make movement and moving stuff and looks are the main reason people use SL, right? The need to chat can be handled by text through IRC or forums et al. Not only that, but even voice or video fill peoples need to interact with a human. So, the fantasy and visuals help out and the better they are the more you can express or emote from the experience and all within the shell of fantasy so you can step away and be removed to aid in healthier mental health BUT be engaged, unlike TV where you are more passive. Maybe it is all bad though, maybe we will end up total SL junkies and all be on TV as they remove the wall from our rooms to pull us out and load us up for emergancy liposuction. OR, we end up super skinny supermodel types due to not wanting to leave the PC to eat! I figure it all just makes it more fun, so I will paly around with it and excape my increasingly upsetting life and watch shadows roll around as I move the sun with a slider and dream a little about the neat mechinima I could make if I had a better PC OR find one of those TV in cards and use the svideo out on my present PC....which reminds me! I need to do RL stuff...wow, I am time wasting agian.

[DanielRavenNest Noe]

---

Poenald Palen wrote:

SL is moving toward becoming a leader in mechinima, IMHO. The camera control, the lighting and now features like DOF are all very much stuff that is going to make much more artistic stuff.

---

Sorry, but that made me smile.  If you want to see what a real leader in cinematics can do, take a look at the "Track View" pages from the Cryengine 3 manual: http://www.mediafire.com/?pkbudm63bai4mu9

 

[Naiman Broome]

Heheh we love the Cryengine don't we ? ....

[Feynt Mistral]

I'd like to see procedurally generated textures myself.  It's a lot more complex though than "hey, there's an engine, let's use that for our texturing!"  Windlight for example took a long time to get implemented, and they even brought in members of the team that made it to work with the people who "knew" the SL client (could anyone really know the old SL client's code base though?).  I feel an important step toward supporting procedural textures is the planned material system.  Not only will it allow conventional texturing which we're all comfortable with, it lays a solid framework to put the aforementioned Substance engine into (if they choose to go with that).  Not only that, procedural textures are less useful than the textures you would draw yourself.  They're designed to fill an area with interesting stuff (like grass, tree bark, or craggy rock.  Maybe a dungeon wall hewn from blocks of stone), but they don't get more specific than that.  User defined textures can do that (with outside procedural textures, maybe) and more.