http://renderman.pixar.com/view/latest-tech
1st video when you scroll down. All those beautiful sliders about 4:15 in. Can it get any more intuitive and faster then that. Vicky have you seen this? A render engine that doesn’t get between the artist and the work. Could it be any less intrusive then this. Someone has put a lot of work into this with the end users in mind.
Check this video from Sig 2013!
What’s so special about that? It’s just an UberShader.
Which is great and would be good for Cycles instead of using 10 nodes just to make a glossy material, but it doesn’t look anything special or new. UberShader is sort of a ‘given’ requirement, it just hasn’t been seriously coded into Blender yet.
There are node groups which do this but the interface is not efficient due to limitation of that method (there’s about 50 lines of properties with no toggles etc) and is probably not efficient to render.
no? takes extra space? now sliders in blender are perfect.
look, I use 3delight, not renderman, so I do not know the second (but the guy who writes shaders and scripts for 3delight said renderman is more complex).
What I can tell you is that 3delight IS NOT easy to use. To make, for example, a refractive material is not enought to add a node. You have to add special lights, write the right names in the right slots, connect the nodes in the right place (and is not somthing like the material output of cycles).
You can say a lot of things about cycles, and for sure an uber shader would be nice. But compared to 3delight (and I think renderman as well) its usage is a piece of cake.
I agree, We used 3DElight for several years before going Arnold, and although I prefer 3DElight render over Arnold in some aspects, Arnold is a breeze to setup, except maybe for interior scenes!
Isn’t that what Renderman’s new alternative RIS renderer is supposed to help with? Providing a simpler physically based renderer that is more comparable to Arnold and V-Ray that might be more suitable for a less technically focused production team.
Yes, as demonstrated in the video I posted, I was referring to 3DElight in my previous post!
Personally, I don’t see the appeal of an ubershader. say what you will about the complexity of adding a handful of nodes to do a compound shader, but having a truckful of sliders glaring at me when I’m only doing a basic diffuse is obnoxious. Add as many sliders as you want, there still will never be the same level of control that you can get with nodes. Nodes aren’t immediately obvious, they take some time to learn, but so does UV unwrapping, texture painting, animation, and pretty much everything else in CG. I will just never understand the node-phobia.
I strongly agree with you
I like the concept of a multishader or two, that combines several common tasks (like in the video example) while making sure physical rules are adhered to. Then noobs (like myself) can focus on other things rather than study how materials behave and leave that for later - it will come in due time anyways usually. But these multishaders should reside as vanilla shader types (just as glass is a combo of refraction, gloss, and fresnel). I suspect this level of programming offers more than what can be achieved through node groups. Even the math node could use some additions to help declutter the node screen (abs, lerp, smooth(er)step, weighted averages, polar transformations and so on). I’m not talking about replacing the node tree, rather than expanding the possibilities.
Ubershaders are for noobs, right? Basically the Blender Internal render material is an Ubershader. The real power comes in being able to just throw out the parts you don’t need to recover some speed. All those nodes or shaders in the background do come at a render cost.
ubershaders are for noobs like procedural textures are for noobs. we’ve all seen terrible examples of hastily applied wood procedural textures or the ubiquitous checkered floor.
There are also lots of really powerful effects that can be done with procedurals, but if the only reason you’re using procedural textures is because you don’t know how to use uv mapped image based textures, it will probably show through in the end product.
and indeed the real power of 3delight (and renderman) is the ability of customizing EVERYTHING. Not only you need nodes to build a material, but you need to write them and assemble them is not easy. I guess they are showing that ubershader because they are trying to attract a less high end kind of user base, as the price revision shows. Don’t get me wrong, I am not saying that if someone uses an ubershader is a bad artist, but I do not think in pixar they use ubershaders 
That is not entirely true. They are useful.
ubershaders are for noobs
That’s not true either.
Have to agree that nodes are the best once you get used to them.
I took a bunch of classes on MAX/MSP a long time ago (it’s a node-based audio system that can basically do anything you want with sound/music) and using Blender nodes was very easy once I understood that.
I think once you get the idea of information being passed around from outputs to inputs, it all starts to make sense.
If you are used to the paradigm of unitarily constructing a material, texture, or light, and then individually applying that to objects, nodes don’t make as much sense.
That said, nodes are great once you get them. They offer almost as much flexibility as coding your own stuff, but are visually much easier.
Ubershaders are for noobs, right?.. The real power comes in being able to just throw out the parts you don’t need to recover some speed. All those nodes or shaders in the background do come at a render cost.
Ugh. If a tool makes a job easier or quicker anyone who’s not an elitist noodler will be all over it.
If spending 2 minutes to make a glossy shader setup makes you feel pro then great, but most people like to just make materials in the quickest and easiest way possible.
A properly implemented UberShader would not be so stupid to rendering un-needed components.
Nodes are great for flexibility and complex materials, having an UberShader doesn’t mean you lose that ability. It just means for simpler materials you can make them much quicker and easier.
Most materials are not much more than a glossy material with a texture or 2, UberShader for that is great.
Having UberShader does not mean losing nodes.
Just like Vray and most of the other render engines in the world offer an UberShader for artist efficiency… all those noobs.
Anyway something like this needs to be done at least for glossy materials. Right now we can’t even make an accurate glossy material. The IOR/ fresnel /roughness thing is not done correctly and with high roughness the result is very bad with a rim-lit fuzzy like appearance.
Because IOR is not connected to roughness, well, that is, without an extravagant setup (to the point you might as well just code your own render engine) (like a pro would:rolleyes:).
@Rhys: “Just like Vray and most of the other render engines in the world offer an UberShader for artist efficiency… all those noobs.”
Took the words right out of my mouth.
in cycles it’s common practice to make your own node groups which gives you a reusable modular uber shader… you get the best of both worlds…
whether that’s a simple @fresnel" (mixing glossy, diffuse viea fresnel node or whatever…) or somthing more complicated liek a car paint or skin shader (or even a BI replication)
it’s noobish to make sweeeping statements… like uber shader’s are for noobs…
generally if i do something more than once on different materials I’ll make a group for that. If i’m sensible I’ll use a group that has the functions I need and there’s no speed penalty…
by teh way the renderman stuff only exposes the shader controls that you set it too in teh shader…
and if teh shader is written sensibly it can be combined in node groups too…
renderman is traditionally more hardcore than using nodes as you need to write a shader (used to be using rsl… not sure what RIS uses for shader definitions…
so in maya you’d be getting teh best of both worlds: pick a shder for your materials and use the sliders and controls that shader offers, make a node set and work like in cycles… or mix and match.
Nothing to see here, nothing “revolutionary” just don’t see the point of this thread…
That said a “switch” node in cycles would make node groups better… expecially one that couls activate or disable inputs on a node group…
but isn’t that moot from a performance basis these days since Brecht made unused nodes not get calculated? just a ui ledgibility issue…