I’ve never had to work with PRman outside of personal stuff, so my opinions are void. However, what you (OP) are saying about troubleshooting time vs actual lighting (or fx in my case) is something I do find myself frustrated with on my latest project. I’m not master of my craft yet and I think theres alot of stuff I see others doing which is helping them work cleaner and smarter. However, man, I do find it frustrating when you spend 3 hours having a blast setting something up and its done, ready to go… BUT WAIT… the motion blur creates huge streaks on certain frames, or one element seems to have a weird shuddering, or the farm kicks back generic errors like “exception found”. I think 90% of it is probably my stupid fault too… but lately I have noticed I spend a hell of alot longer troubleshooting my FX than actually making them.
Renderman is dead!! Or at least on its way to die!!
Renderman tried to incorporate modern raytracing features and adopted workarounds & workflows that made it the black art of todays CG world. Those who grew up with it swear by it but youngsters don’t hesitate to call the king naked.
“Big” studios will continue to use renderman, but there are less “big” studios every year. I’m not sure if Arnold will be the answer for film work which relies heavily on moblur/dof rather than physically correct reflections or such but “realtime” gpu renderers will eventually replace everything else in the long run IMO.
I think Mustique is right. You have the “older” gen of 3d guys who are used to Renderman and forgive its flaws because it gets the job done everytime, coming up against a newer generation of artists who grew up and learnt cgi using MR, Vray etc…
I see both sides. And its not just in render terms that this is happening. Older edit machines, colour grading systems, compositing tools that are still installed in many studios are stil being used daily even though there may be better systems because they get the job done . Things are changing from what i read here on cgtalk, just slowly, since the most important thing for us all is getting delivering the shot on time. And if that means not using the most up to date, fastest tech, then so be it.
And if that means not using the most up to date, fastest tech, then so be it.
Frankly, if you need to render 100 furry creatures with motion blur/dof, Prman will probably do a better job than any ray-tracer (Vray/Mental ray at least). So I would say in that case it is the fastest/most up to date tech.
(just to put my cards on the table, I love ray tracers, just don’t think they can do everything a Renderman renderer can do quite yet, in my experience at least).
I agree with ndeboar, you just cant beat renderman with all of those taken into account, raytracers are still very slow and will be for a long time, they are great for smaller facilities.
Sorry, i wrote that badly. I shouldnt have written “fastest” when i meant newest. I just meant sometimes the older tech is the most robust and thats the reason its still used. As you said about the 100 furries.
renderman will die when our computers will be 100(0)x faster, yep? It’s all about processor speed&memory…No more jupiter-jazz services, no more shader tds, just throw everything into brute force raytracer with ubershader and wait a while for a result…so many ppl out of job 
If you worked on guardians to achieved that level of realism with feathers
… I am amazed at your work. You did a fantastic job.
Keep in mind that for dramatic CG (full 3D features, shows, and the like) the directability of the light is often even more important than the accuracy of the light. So when the director instructs the lighter to drop the intensity of object A, and make object B’s specular softer on the face but not the body, and to put a specular hit on C’s eyes so they can read better even though there isn’t a light there, and to remove object D’s shadow but keep the object itself visually in the scene, then you really don’t care what is physically accurate. You care about what you can tweak or fake in order to get the shot out the door.
Also when you have to render those 100 furry creatures running across the furry (grassy) ground plane with thousands of rocks and pebbles scattered on the ground as mist hugs the ground and smoke rises from the burning forest behind them… there is simply no way you can do it all in one pass with any renderer. So you need the flexibility to render the ground plane with shadows from the characters (but with stand-in fur since full fur takes too long), and then render the characters with the grass and ground cutting mattes into them (but only the area around the characters, since using the entire terrain’s grass is too expensive), etc.
For complex shots your renderer needs this kind of flexibility and complexity. You need to be able to customize the shaders, lights, surfaces, and parametric geometry. And you need to be able to fake stuff both for time and flexibility because you really can’t afford to render the background over and over again, or calculate shadows or GI exposure over and over again.
Do VRay, Mental Ray, Arnold, etc. each have this kind of flexibility? I haven’t used them, so I can’t say. But I have used PRMan in production before so I do understand why it is structured the way that it is.
Cheers,
Michael
Not so verry long ago most ppl on these forums were convinced raytraced GI was never going to be used in film either because its too slow…
I also think over time raytracing will completely take over, just because its so much more userfriendly (less headaches and no more losing hairs!).
Nothing against renderman, its done and still is doing some beautyfull stuff.
Never…No Comparison, it will never gonna die only because of slow ray tracing or because comps are getting faster…
when it comes to handle THINGS LIKE motion blur, DOF, displacement(with almost no extra cost of render time), programmable shading for films with hundreds or even thousands of characters with fur, hair, cloth within a complex environment, I don’t think there is any renderer that stands here. most of the times you have to do fuzzy things in order to get directors vision within the given time and budget. In my opinion Renderman is the only renderer that is production proven and the Reyes pipeline is one of brilliant invention at the time because it was designed for films in mind and designed such way that even today it is able adopt point cloud technique after almost 20-25 years.
Think about the King kong’s fur, think about the complex world, environments and characters in Avatar, I don’t think any of these renderers can do it reliably.
Any ray tracer in todays market ( haven’t used Arnold) like MR VR Maxwell or any… they simply not gonna compete with Renderman because of its - RELIABILITY, FEATURE SET, FLEXIBILITY, ROBUSTNESS, PERFORMANCE AND WONDERFUL IMAGE QUALITY.
I think We all must Respect The Renderman and its Specification who showed us the wonderful world of effect movies like Jurrasic Park, LOTR, Terminator, Transformer, Avatar…and not to wait for its end.
Doing some render TD work from time to time I wait for this moment to happen. My approximation on when this will happen is - NEVER. Since, requirements and complexity of VFX rises with added bonus computer power each year, the render times stay the same and the task of optimizing the scene and coming up with clever ways to render it is always in place for the past 30 years, we will never go bruitforce! Soon they will decide to do movies at 8K in stereo at 120fps - and there goes your bruit force out the window =)
Flexibility, scalability, creativity are important. How do you render smoke, ice, fire, iridescence, and a hundred other things in a directable way? Those kind of things need creative solutions. And how about rendertime generated geometry, pulling data from file, and such things? A shader isnt just a representation of a physically correct process. You want to incorporate artist specific things and optimizations. What if you want to create shadow maps once and reuse or distribute across a farm? Create AOVs? With mantra/vex you can even go the other way: manipulate your scene geometry with the shading language (yes i know, not quite the same, but reyes and the language to create the shading exist side by side so i believe the argument is valid)
We cant wait for oneclick solutions and thats why things like renderman are her to stay for a little longer.
Why is this and argument about reyes vs raytracing? Renderman does both of these things and it seems each version the raytracing is improving… Both obviously have their place and situations where they excel…
PRman as a product and its position in the market place has got to do with many factor not just technology, it wouldn’t be far fetched to say PRman is loosing some ground these days but that is due to many factors…
The bigger facilities are raytracing and brute-forcing more and more everyday. (some)Raytracers, when the total amount of hours on the farm for iteration are counted, can already beat rasterisation in many cases.
Every major shop I know someone in is in the process of, or is about to start, reworking or adapting their pipeline to accomodate more raytracing, uber-shaders, energy conservation models, iteration friendly methods, and reducing explicit front-load computation from its current all-time highest to something less ridiculous.
PRMan, like many other products, is addressing the above trend because:
A) Pixar needs to for their internal productions
B) That is what clients are asking for.
If you think PRMan and raytracing are mutually exclusive you might need to take note of the release notes for v16.
The idea that rasterisation and reyes will remain the predominant forms of bucketing and pushing those buckets is mis-founded and mis-informed, as is the notion that a forward path tracer isn’t controllable in the details at all (IE: regarding some of the examples given, you can implement inclusion lists and artificial speculars just fine and let the energy conservation model only deal with the diffuse, or not at all)
The truth is in the middle, we’ll still need both options for a while. Renderman as a standard might or might not be there for all the while, but then I think the OP intended PRMan.
PRMan is doing its best to change its skin, retain the pipeline and dev friendlyness, but not be such an antiquated sloth anymore, so it might very well survive the transition and reposition itself once again as the market’s gnomon. Or it might not
As I’m just an indevidual, I can’t speak for RenderMan or the REYES pipeline… but this is how it’ll probably happen.
Computers will get faster and faster… but then someone will (if they didn’t already) come out with a nice new set of shader algorithms that make thigns look SO MUCH more realistic than they do now… and render times will be just as they have before… s that being said, I don’t think we’re ever going to hit a real end point to where raytracers will be quicker.
I found an applications that uses the GPU to help raytrace are still pretty slow.
Though I love Vray and such, when I was at Method they went from being a Vray/Mantra place (with a little bit of renderman going on) to using PRMan as their main renderer. It kind of shocked me, but it just shows you, PRMan isn’t going anywhere…