Mental Ray Interface

[peliosis]

I was wandering will it ever be possible to have a direct access to MR (like XSI rendertree) from within VOPS or rather specialized interface.

Mantra is everything but a fast raytracer:), and writing shaders for MR is far less usable than laying VOPS nodes.

Peter

[TheUsualAlex]

Out of curiosity, isn't MR shader entirely done with C/C++ extension? I supposed that, as usual, if there's enough customer demand for MR VOP, I am sure SESI could do something.

[kodiak]

Out of curiosity, isn't MR shader entirely done with C/C++ extension?

mr has 2 ways to define shading properties:

1, shader dso: functions written in c/c++, published as dll and linked to the mental ray core rendertime.

2, phenomenon: a text file describing a shading tree built from the compiled dlls (much like in maya hypershade)

On a VOP level both would be possible, but probably most users would be interested in the second one, having a nice shader base, and just wiring them together much like they do in XSI.

I think this doesnt really need much developer effort, as the phenomenon syntax is really simple, it doesn't have control structures and such stuff that would make parsing a tree nasty. Well defined ports passing data around, and mr comes with a more or less satisfying shader library (the rest can be chased down from xsi/maya sites anyways).

Personally I think (altough i don't really use it recently) mr is getting there, so probably would be a nice addition (of course there are other priorities as well).

[andrewc]

Mantra is everything but a fast raytracer

You should give mantra in Houdini 8 a try. There have been some significant improvements to the ray tracing performance. For some scenes I'd say its faster than MR with all other quality settings being equal.

[hoknamahn]

mr has 2 ways to define shading properties:

1, shader dso: functions written in c/c++, published as dll and linked to the mental ray core rendertime.

2, phenomenon: a text file describing a shading tree built from the compiled dlls (much like in maya hypershade)

On a VOP level both would be possible, but probably most users would be interested in the second one, having a nice shader base, and just wiring them together much like they do in XSI.

I think this doesnt really need much developer effort, as the phenomenon syntax is really simple, it doesn't have control structures and such stuff that would make parsing a tree nasty. Well defined ports passing data around, and mr comes with a more or less satisfying shader library (the rest can be chased down from xsi/maya sites anyways).

Personally I think (altough i don't really use it recently) mr is getting there, so probably would be a nice addition (of course there are other priorities as well).

20801[/snapback]

I don't think that first way is easy to implement because developers need to:

1. Implement MR's function library as VOPs

2. Write exporter from VOPs to C++

3. What about shader preview (for that C++'ed dsos)? On the fly compilation?

And other thing... This way requires installed C++ compiler. For *nix users it's not a problem, but for Windows users? Certainly there is nothing impossible

[kodiak]

I think for a start a phenomena implementation would be enough.

From my experience with mr in maya and xsi, you don't really need to write c++ code if you have a nice shader base (and you do). And if you do, you better start in a C++ IDE anyways, cause it will be a pretty general, reusable component that's worth generalizing and optimizing, debugging..etc.

There was (is?) a tool called Shadetree around, that was able to create shaders from visual trees much line VOPs do, and had support for prman, vmantra and mr as well.

Cinegrfx - shadetree

[peliosis]

Seems like there is very little interest in MR for houdini subject.

Does any company use them togrther to some serious extents?

If it's not too hard to write such an interface I wonder why (or if:))

somebody haven't wrote it yet?

Maybe big people rely on hardcoded shaders and don't care for any conveniences.

Maybe if MR would be added to houdini as it is to some minor apps like max;) people would be more eager.

If I had ANY knowledge of some C++ and other blackmagic...:):)

cheers

Peter

[Jason]

Seems like there is very little interest in MR for houdini subject.

20941[/snapback]

There seems to have only been MR fan interest since the days of Crescent Moon Productions which inspired the development of the .mi exporter in Houdini as it is.

Now that VOPNET_Node has been in exposed in the HDK does this make it possible to create one's own VOPnet types? If so, a very eager person could tackle a phenomenon shade-tree for Houdini perhaps?

[Guest xionmark]

Hi Peter,

It's true that MR is probably the least used renderer utilized by Houdini shops, but it is used by a few, and very impressively I might add. I'm new to MR and will be working the next few months here at Charlex to deploy MR with Houdini and will be glad to post any interesting tidbits and tips that I may discover. It does seem to generate some mighty fine .mi files, but again, I'm just getting started.

It would be very cool (and I think not all that difficult) to have phenomenom networks created in the VOPs editor. But since I haven't done it yet, I may be wrong.

Cheers!

--Mark