Bear with me, as I'll get to my question after some set-up using a comparison between MToR and Mental Ray for Maya... also, the following has been my experience/understanding, so any place it says "I know" you can feel free to replace it with "I believe" if my info is wrong (if it is wrong, it will probably be most wrong in the MToR sections, as I have limited experience with MToR). Ok, let's begin: I know MToR generates a RIB file, which automatically renders out using Renderman. You can tell it so save this RIB file for later use. I also know that you can have Mental Ray for Maya generate a .mi file, but, according to the MR for Maya documentation, this .mi file can only be rendered using the standalone Mental Ray renderer (which costs roughly 2 grand, I believe). I also know that both MToR and MR for Maya need to translate your scene into something which can be rendered using their respective renders (Renderman and Mental Ray). If you have an animation you wish to render, MToR will translate the whole scene once (from the start point of your animation to your endpoint), generate the apporpriate RIB file (which you can choose to save to disk or not), and begin rendering. I would imagine that MR for Maya, like Renderman, generates its own file, a .mi file, whether you tell it to or not, and if you tell it to not save it, it just trahses it when the rendering is done, a la MToR. What I don't know is, for an animation, whether or not MR for Maya translates your whole scene and begins rendering, or if it translates on a frame-by-frame basis (I've rendered animations a couple of times with Mental Ray for Maya, but to be perfectly honest I always left the render going while doing other things, so I didn't notice if it kept translating for each frame, and instead of testing it right now, since I need to get some sleep, I am just going to ask the following question). So, does MR for Maya re-translate the scene for every frame? Even if it doesn't, translation is still an issue. A colleague of mine has a relatively high-geometry model that he has been doing render tests with in Mental Ray for Maya, and it takes roughly 25 minutes just to translate the scene... granted, his scene is not terribly optimized, but still... thats a long translation time. Imagine if you were doing an animation with that, and somewhere in the middle it stops for whatever reason. You'd have to have it re-translate for 25 minutes (or, easily, much longer if your scene is more complex) just to begin again at whatever frame it left off at... You can't just open the .mi file and tell it to render from whatever frame 'til the end (as you can do with a .rib file using MToR/Renderman), you need to have it re-translate from that frame. If you have the standalone Mental Ray, I am sure you can use your .mi file and tell it to start at whatever frame, but, even though there is a copy of Mental Ray with MR For Maya, it won't read .mi files, apparantly... only stand-alone Mental Ray will... so we are looking at lots of translation time any time you wish to continue rendering an animation from the middle, or whatever... now, I do understand why it needs to be translated, since it needs to go from a Maya file to a Mental Ray file, however, XSI, if I am not mistaken, needs no translation to render with Mental Ray. Its default renderer is Mental Ray, thus its objects and shaders and everything are inherantly Mental Ray files (its also renders a whole lot faster, apparantly, but they've had quite some time to work out all of the issues with implementation, whereas Mental Ray for Maya is still in its infancy). Also, since Mental Ray is XSI's default renderer, you can tell XSI to re-start rendering from the middle of an animation, and it won't have any problems (such as long translation times). Anyway, I guess this turned into more of a rant about wanting MR for Maya to be better, rather than any real question... I know it is still an early, early, early version of the software, but I think Alias|Wavefront should take a cue from Avid/Softimage and simply kill their renderer, and have all of Maya's objects and shaders just be Mental Ray objects and shaders... well, they can keep their renderer for now for the express purpose of rendering fur, paint fx, etc, but since mental ray is vastly superior in every way shape or form, they should implement it FULLY for everything else, thus eliminating translation times. Is this completely unrealistic? Probably. Will it happen at some point in the future? Perhaps in a nother 2 or 3 versions of Maya... but, hey, a man can dream, can't he? Anyway, if any of this post/rant is woefully innacurate, please respond telling me that I am an idiot. But also please include why I am an idiot and which specific parts of this post are wrong and thus make me an idiot. Thank you for your patience.
-Steve
p.s. Let me apologize specifically to those of you who have read any of my other posts. For some reason I can never seem to write a short post...