Best Rendering Engine

[WireFX]

Renderman? Renderman what? Renderman itself is only the name of the SDL standard but no specific renderengine. In the case you mean PRMan you should write it. Otherwise this could also be BMRT, Air, 3Delight, Aqsis, Pixie, Entropy or even something more exotic.

PS: As far as I heard the most realistic renderengine should be inside of RADIANCE
PPS: An interesting comparison can be found here: http://www.winosi.onlinehome.de/Comp1.htm

[ZeBoxx]

Just out of curiosity - why the limited set of options (short of whatever the CGtalk poll limitation may be) ?

There's really a plethora of renderers out there.

The ones you mentioned are all available for max (renderman through MaxMan), of which two are specific for max, so I'd imagine you might be aiming at the MAX market specifically. If so - I'm not seeing V-Ray, Raymax, virtualight max (exporter to virtualight, really), etc.

I think the scope of this poll is a bit too small to really get a statistically good result.

Justy my €0.02 :)

[fxjeane]

having a complete shading language that is both easy to use and expandable (through shader DSO's) makes PRman the most robust renderer on my opinion. You must also consider that it is compatible with ANY animation/modeling package through a 3rd party plugin , or throgh your own custom code.

With PRman and the RenderMan Standard you get an open set of comands to which you can plug into to produce some very wicked stuff..

THAT is the reason why it is the MOST used renderer in the VFX/animation industry.

Mental Ray seems pretty good, but you must be even more of a "Propeller Head" to tap into its true power.

cheers
FxJeane

[Manifesto]

ZeBoxx: Yes it was limited to the # of options I could have on the poll, those are the most popular I know of. I understand there are more out there.

WireFX: Thanks for the link. Is there anything else out there like this that use the same file and render it with engines like brazil, mental ray ect..? I would like to compare with a more consistent image.

fxjeane: Thank you thats the kind of response I was looking for. I know people use different packages w/ different rendering engines. But There must be one that is more consistent among others for high quality results.

Basically I'm working on a demo and I want a realistic quality to it, if there are any suggestions pleace feel free to let me know.

[Oktavian]

Isn't Mental Ray still the industry standard?
Therefor its a strange result until this point :) .

[LeeGriggs]

lets not forget hardware renderers like PURE for Maya and Max:

www.artvps.com

[Qianniu]

I like Bireazl!!!

[fxjeane]

Originally posted by Oktavian
Isn't Mental Ray still the industry standard?
Therefor its a strange result until this point :) .

I doubt that there is an "Industry Standard" That was what the Renderman Interface was supposed to become, but it never quite materialized. With the advent of new RenderMan compliant renderers we might be heading in that direction.

I doubt that mental ray is more used than PRman in the film/vfx industry, all though it is gaining a lot of users since it was bundled with Maya.
The problem is that a lot of people think that because they can render good images with Maya/Mray that they "KNOW" Mray. Mental ray is deeper and harder to use than PRman. It's like people who know how to use PRman through Lightman or MtoR and think that they know PRman.... On both of these packages you need to know the API to REALLY know how to use them.

Fxjeane

[Manifesto]

brianbriggs: I looked at the rendering engine PUR, I saw some examples on your website. It looks really good. How did you find out about it? Seems to be better for objects like cars, buildings, desks ect.. rather than human characters. Would you agree? If you were to model a human would you use the same engine or a different one? If so what one would that be?

quote:
--------------------------------------------------------------------------------
Originally posted by fxjeane
The problem is that a lot of people think that because they can render good images with Maya/Mray that they "KNOW" Mray. Mental ray is deeper and harder to use than PRman. It's like people who know how to use PRman through Lightman or MtoR and think that they know PRman.... On both of these packages you need to know the API to REALLY know how to use them.
--------------------------------------------------------------------------------

What do you mean by this fxjeane? Please explain.

[LeeGriggs]

>Seems to be better for objects like cars, buildings, >desks ect.. rather than human characters. Would you >agree?

I believe their image gallery is aimed at the differing professional markets - such as ind. desgn, architecture and automotive. That doesnt mean it isnt capable of rendering people, its just I imagine their isnt as much money in the rendering 'humans' market as there is in say architecture or design.

But the main strength in using a hardware renderer such as PURE is its speed, and my god what speed. :eek:

I am regurlarly getting speed performances that far outweigh what my pc can do. If you check out their speeds on the website, you can see what I mean:

http://www.artvps.com/products.ihtml?page=comparisons

Not only does it speed up my renderings, but it also speeds up my workflow, which is invaluable.

[fxjeane]

Originally posted by Manifesto

--------------------------------------------------------------------------------
Originally posted by fxjeane
The problem is that a lot of people think that because they can render good images with Maya/Mray that they "KNOW" Mray. Mental ray is deeper and harder to use than PRman. It's like people who know how to use PRman through Lightman or MtoR and think that they know PRman.... On both of these packages you need to know the API to REALLY know how to use them.
--------------------------------------------------------------------------------

What do you mean by this fxjeane? Please explain.

What I mean is that most High end renderers (like Mray or Prman) are stand alone renderers which generate images either through API calls or with their own internal file format (PRman's RIB).

When you use a plugin like the Maya/Mray or MtoR and don't bother to learn how the renderer really works, then is very unlikelly that you will be able to take complete advantage of the renderer. You can build nice shaders with MtoR slim networks, but what if the effect you are after needs some funky BRDF and slim doesnt have a built in function for what you need. Thats when you roll up your sleves and write it yourself. You will not be able to do this unless you know the renderer.

I dont think that plugins support ALL of the features of these renderers, and if they do, they will do it through some open doors (like MtoR's RibBox) where you can insert you own code.

Its a delicate matter because a lot of people that use Mray with Maya feel capable engough (and im sure a lot of the are very, very good) to say that they know Mray... and I just feel that these migh be a little missleading......could they use Mray to the same extent if they used it in XSI? It will probably take them a while because XSI's Mray connection might be completely different from Maya's. If they DO know Mray, they would adapt a lot faster.

cheers
Fxjeane

[Manifesto]

Thanks for clearing that up fxjeane. For people that want to REALLY LEARN things like maya PRMAN, MRAY and MtoR what do you recommend? Do you have any references, sound advice? Thanks. :cool:

As well, anyone, what rendering engine do you recommend for rendering organic substances? Like human skin?

[WireFX]

At http://www.cgtechniques.com/ are also some good comparisons between different renderers though they are almost architectural renderings.

[Mauritius]

As there was a lot of discussion about PRMan, I feel someone should note that there are RenderMan compliant renderers which are very close to PRMan and some of which are free:

3Delight (free)
AIR ($400/CPU)
Aqsis (free)
BMRT (free)
RenderDotC (price varies)
Pixie (free)

I'm too lazy to fiddle all those links, use google and you'll find all of the pretty quickly.
For starters, read the book 'Essential RenderMan Fast' by Ian Stephenson.
Then move on to 'Advanced RenderMan -- Creating CGI for Motion Pictures' or ARMan for short.
Visit news://comp.graphics.rendering.renderman often and don't be shy to ask.

Cheers,

.mm

[jeremybirn]

Originally posted by Oktavian
Isn't Mental Ray still the industry standard?
Therefor its a strange result until this point :) .

Most films that came out this year, or are currently in production at big effects studios, are rendered in Renderman. But, many of those studios also have and use MR as well for some effects, so yes, it looks as if MR is growing into a sortof "second industry standard" in way.

The popularity of Brazil on this pole is just because it is a Max plug-in renderer that is affordable and easy to use, not because it has broken into a lot of high-end companies yet. Even though it hasn't yet, and doesn't have an interface in Maya, it still could grow and get installed more places over the next few years.

-jeremy

[jeremybirn]

Originally posted by Manifesto
For people that want to REALLY LEARN things like maya PRMAN, MRAY and MtoR what do you recommend?

Start by geting a good math and computer science background. Programmers who want to go far beyond the GUI need to understand the math that makes things work under the hood. Even to read a SIGGRAPH paper describing a new rendering technique and understand it, you need some advanced math.

Although, a lot of people get good results without writing any new shaders. In terms of individuals working alone, the people who get really good at other skills (like painting excellent texture maps) tend to render better scenes than the ones who spend all their time on shader development.


Originally posted by Manifesto
As well, anyone, what rendering engine do you recommend for rendering organic substances? Like human skin?

One of the big issues in skin rendering is getting correct "Sub-surface scattering" for realistic translucency. There have been some long threads about that already:
http://www.cgtalk.com/showthread.php?s=&threadid=50332
but if you can get a renderer like Brazil that has some support built in for SSS that's a huge plus.

-jeremy

[Apoclypse]

I agree with the PRman statement and it robustness but you forget that many studios are now switching to Mental Ray because of its extensibility. Mental Ray can be used just like PRman even to the point that you can write an MR shader that translates a PRMan shader. Writing MR shaders is quite easy and that is becuase of the language ( c,c++). The RM shading language is great but sometimes requires quite a bit of coding as compared to MR, which can have shaders that written more tersely). MR can be used with any program out there ( writing a parser for MR can be done the same way as RM after all they both use a file format that can be written and read through a text editor so it stand to reason that they can be scripted to.)

MR is a great rendering engine and like PRman can be used on almost any app and is not limited to one extremely buggy app. (MAX) I noticed you guys put brazil at the top of your list but i beg to differ the most powerful renderers should work around your project not your project work around it. If you were a big name studio or even a boutique would you choose a renderer that is tied to one app or a more versatile renderer with the ability to write your own shaders without having to worry if it will work with a certain app or not ( becuase it was written for the renderer only)

[Mauritius]

I agree with the PRman statement and it robustness but you forget that many studios are now switching to Mental Ray because of its extensibility.
What studios specifically are you talking about? Where is PRMan less extensible than MRay?

Mental Ray can be used just like PRman
even to the point that you can write an MR shader that translates a PRMan shader.

Where would the point be in doing so? Why not either write a C++ version of the shader or use a RMan renderer beforehand? Production shaders can be several thousand lines of code. If you have a shader that you want to re-use for just one sequence and you switched your pipeline to another renderer it may indeed be cheaper to buy a few licenses of a RMan compliant renderer (not neccessarily PRMan!) and re-use that RMan shader instead of having an expensive TD sit down and try to come up with a MRay version based on the RSL source (which may well take several weeks, for a complex shader)

Writing MR shaders is quite easy and that is becuase of the language ( c,c++).
Lol! You apparently never wrote a MRay shader -- at least not for an actual production. Nothing could be farther from the truth. Search the highend3d rendering theory list archives for some comments on this very topic by shader writers from ILM to get an idea how wrong you are.

The RM shading language is great but sometimes requires quite a bit of coding as compared to MR, which can have shaders that written more tersely).
This is nonsense. RMan SL (which is basically C,) is a much easier way to write a shader. Anyone who ever had to do this for both renderers will wholeheartedly agree. The team at ILM back in the late 80's simply found the -- as of today -- best compromise between features and complexity. While in C++ you can do anything, in a shader you don't want to do anything, 90% of the time. A langauge which was designed to support a specific task is most often preferrable to a general purpose language (interestingly this is also the reason why higher level languages exist at all -- otherwise, why not write everything in assembler using a hex-editor ;) ). This is in particular true for writing shaders.

And another reason people seem to forget about a RMan based pipeline (vs. a MRay based one) is simply that: choice of renderer. If your pipeline is MRay based, you can only use MRay. If your pipeline is RMan based, you can select from a bunch of renderers that adhere to this standard and which range from being completely free to the $3.5k PRMan (yes, Pixar lowered their prices considerably).


Cheers,

Moritz

[fxjeane]

Very well put Mauritius.
I agree with you all the way. The RenderMan Standard ROCKS and with the release of new, cheaper and higher end features (GI, irradiance, Hybrid raytracers/scanline, etc) renderers, it will only get better.

I have no doubt that we are not far from seen the first movie rendered with a RenderMan renderer other than PRman (maybe AIR ,3Delight or Pixie).

Learning RenderMan is a great time/career investment for any TD.


cheers
FxJeane

[Mauritius]

I have no doubt that we are not far from seen the first movie rendered with a RenderMan renderer other than PRman (maybe AIR ,3Delight or Pixie).

Actually, RenderDotC has numerous feature film credits already. I believe Cinesite owns a bunch of licenses of the renderer. I also believe/know that 3Delight & AIR are next ...

.mm

[Manifesto]

Thanks jeremybirn, I visited the thread you mentioned. It was very informative. It is so refreshing to see someone who is so knowledgable of their craft. Inspiring:)

Brazil or Renderman seems to be the way to go in the way of sss supported rendering systems.

[Apoclypse]

Muaritius when I say that writing a sheder n MR is easier to write than RM it is from personal experince and really has to do with my own opinion. I actually use aqsis and BMRT a whole lot and have written quite a few raymarchers for both and I still find that wrting one for MR is far easier and far faster than in PRman. Maybe its just me. Yes it is true that some things that will work in RM won't necessarily be easily reproduced in MR because of the different technologies being used in both renderers.

However anyone who has a written a plugin for maya will be more at home with MR becuase they both use the same language.

Obviously you have had no experince using either renderers shading language because you would soon figure out that C needs far more code than is neccesary as well as macros and quite a few thing that though are faster (the code will be faster) are far more complex to control than in C++ where a previous 25 line code can be written with far less in C++. Besides SL is NOT C it is however a c based language (there is a difference)

Yes there are a lot of renderers out there that are free or cheaper than MR ( except PRman which is still a little bit more expensive than MR if you get it through softimage). However these renderers range from half assed implementation to that's great but is to slow. BMRT was a great renderer but it was missing nearly more than half of the RI 3.2 Spec last time I checked (and it probably won't ever see another revision again, damn you PIXAR you suck) Aqsis is great and looks to me the most promising yet becuase it is cocnstantly updated.

However they do not have the power or quality of MR in any way shape or form. For one thing like a guy at ILM said (when there was a RM vs. MR post) Mental RAy has is faster than RM by quite a margin.

[Mauritius]

Muaritius when I say that writing a sheder n MR is easier to write than RM it is from personal experince and really has to do with my own opinion. I actually use aqsis and BMRT a whole lot and have written quite a few raymarchers for both and I still find that wrting one for MR is far easier and far faster than in PRman.
Errm -- excuse me but how can you compare PRMan with MRay if what you use is actually Aqsis + BMRT?
BMRT hasn't seen updates since years and Aqsis is not used in actual productions as of now. Besides both renderers are painfully slow.
Additionally, 99,9& of all shaders written in an average production are not ray marchers, so this example says really nothing. Still my reply would be that "Hollow Man", which made extensive use of ray marching shaders, was rendered using PRMan and BMRT by SPI.

Read this post from Ian Stephenson, author of a RMan compliant renderer, a book on RMan compliant renderers, Technical Director (watch the 'Order of the Flame' credits carefully) and lecturer of RMan at a UK univesity:
http://www.highend3d.com/render/archive/sp.3d?mail_id=1973

Maybe its just me. Yes it is true that some things that will work in RM won't necessarily be easily reproduced in MR because of the different technologies being used in both renderers.

As I said -- it's a big differenc if you use a package under production circumstances or if you're just playing around. If you haven't read ARMan and at least the SIGGRAPH coursenotes of the last couple of years on RMan, you will make a lot of false assumptions. In particular by comparing MRay to BMRT + Aqsis.


Obviously you have had no experince using either renderers shading language because you would soon figure out that C needs far more code
[...]

SL is not C. One call to texture() would require 200 lines of C code, if it was plain C and it wouldn't be significant less code if it was plain C++. The reason simply being that this is a high level built in function which does mip-mapping, filtering and antialiasing using correct calculated filter sizes.
RMan SL is simply elegant. Any TD will agree on this. Those guys at ILM had all choices in 1988. They decided to go with a specialzed language rather than a general purpose one. And the fact that e.g. nVidia lately decided the same for CG or SideFX for its VEX, seems to second this decision even more today.

Maybe you never really used SL or read a book about it and therefore erratically compare it to C++ as if it was C.
As I have worked on actual productions with other TDs who use MRay, I know that the same shader in MRAy is usually twice as many lines of code than its RMan SL pedant. A ray marcher may be an exception, but I doubt even this.

[...] Besides SL is NOT C it is however a c based language (there is a difference)

Yes, I said it is "basically C". Not "it is C". Having used this language for seven years to write shaders, and the other (aka C) to do other stuff, I think I managed to discover the differences (somehow, at last). ;)


Yes there are a lot of renderers out there that are free or cheaper than MR ( except PRman which is still a little bit more expensive than MR if you get it through softimage). However these renderers range from half assed implementation to that's great but is to slow.

Sorry, but you again made quite obviously that you are more or less clueless about the topic of this thread. RenderDotC has e.g. been used on Matrix (to name just one example), 3Delight is used in several studios, we used and use it ourselves for an animated feature here to do volumetric shaded clouds via deep shadow maps (even long before PRMan 11 was available with this feature). AIR is also a mature product used in at last two feature film projects and also in many smaller studios (I'm aware of an animated series produced with it e.g.).

Pixie and Aqsis may not yet be on par with PRMan, 3DL, RDC or AIR but they are free and open source. At least where I work we throw our RIBs from actual productions at them from time to time to provide feedback to their authors.
I wonder what implementation you refer to by calling it "half assed"?
Or if you think by how many percent 3Delight or RenderDotC are slower (or faster) than e.g. PRMan to make any statement that regards speed?
And while I'm askin questions to you -- you still have to answer several ones from my first reply, e.g, which studios (you said "many", you remember?) are switching to MRay????

BMRT was a great renderer but it was missing nearly more than half of the RI 3.2 Spec last time I checked (and it probably won't ever see another revision again, damn you PIXAR you suck)
You say BMRT missed 50% of the RI 3.2 spec??? You mean 5%! Even PRMan misses part of the spec.
BMRT has a ton of feature film credits. It can't be so bad then eh?
And what parts (making uop 50%) of the 3.2 spec where missing? -- I'm very eager to hear!
Aqsis is great and looks to me the most promising yet becuase it is cocnstantly updated.
And RDC, AIR, Pixie and 3Delight are not constantly updated???

However they do not have the power or quality of MR in any way shape or form. For one thing like a guy at ILM said (when there was a RM vs. MR post) Mental RAy has is faster than RM by quite a margin.

This is nonsense. First, almost all RMan renderers are micropolygon renderers. If you turn some knobs in MRay to have it create mcropolygon size polygons, it slows down the renderer considearbly. This is never an issue in a REYES renderer.
I reply by quoting what 'gga' from ILM posted regarding the speed issue on the very thread you mentioned and which shows that you apparently didn't read carful enough:

mental ray's scanline renderer is on average about 3 times faster than prman10 and ~1.5 than prman11, when equivalent shaders are used for your average ilm creature (say, our superduper plastic shaders with million of knobs). This is so for preview renders without motion blur.
What DOES make prman amazingly fast during scanline renders is this great fact:

When motion blur is on, performance hit for any prman render is (usually) less than 50% of what it was without.

NO renderer on the market is even close yet to offer this... I know, we usually try them all.
With mental ray (and most other renderers), motion blur implies about a rather constant 500% increase in render times at least. This usually offsets ANY benefits of the mray scanline renderer when motion blur is on.
500% vs 50%...! In those eight years that I work in this business, I never worked on a shot which wasn't ultimately rendered with 3D motion blur turned on.

I think that you simply lack a lot of knowledge about many things you talk about. Just trying to write or writing some shader in MRay or a RMan renderer (or both) is no match to being asked to write a real shader for a real production under real production circumstances.
I have no other explanation why else you could be so off track with your comments.

Cheers,

Moritz

[eYadNesS]

I think Renderman...

[Jozvex]

Just thought I'd throw this article in here:

http://millimeter.com/ar/video_mental_images_industrial/

A quote:

"This strategic partnership between ILM and mental images represents the future of the rendering industry. Studios increasingly demand the level of virtual photo-realism that can be achieved with mental ray software. "

[Mauritius]

Sounds like the average marketing department's newsticker article. Mind you that ILM is the only company except Pixar itself which has access to the PRMan source ...
If you watch the a|w demo reel and have no clue about the industry you might be fooled into thinking the water effects for 'Perfect Storm' where done using Maya. The truth is that a custom plug-in was written to calculate these -- any host app would have been equally well suited probably (Max, Lw, etc.) -- you get my point.

.mm

[Anteru]

Originally posted by Jozvex
"This strategic partnership between ILM and mental images represents the future of the rendering industry. Studios increasingly demand the level of virtual photo-realism that can be achieved with mental ray software. "

Never forget that each and every company says their render is the best, fastest and stable of the world. ILM has been using PRMan since more than 15 years - having developed hacks/tricks which you probably never heard of. They cooperate with mi to have the possibility to use mr if desired. Their main production pipeline still relies on PRMan cause of it's stability and speed. For scenes, where thousands of objects move through a scene, REYES renderers still are fastest/most memory efficent. Just my 2 cents though. RMan4ever.

[Jozvex]

Mental Images didn't write the article though, it's on a news site. Another quote:

mental images, the global leader in rendering software for filmmakers, today announced it has signed a software licensing and custom support and development agreement with Industrial Light & Magic (ILM), a division of Lucas Digital Ltd. LLC, to use mental images's mental ray rendering software for future ILM productions.

"We have had a long and successful relationship with mental images," said Cliff Plumer, Chief Technology Officer, ILM.

And about Perfect Storm, that's pretty crazy because in the book "Maya: Secrets of the Pros" Alex Alvarez (I'm pretty sure) does a tut on how the effects were created using Maya.

[plug3]

It's Habib Zargarpour who wrote about Perfect Storm.
They used Maya but they had to create custom fields to solve some of the problems...

[Mauritius]

And about Perfect Storm, that's pretty crazy because in the book "Maya: Secrets of the Pros" Alex Alvarez (I'm pretty sure) does a tut on how the effects were created using Maya.

Lol! Read the CFX article ...

.mm

[Jozvex]

Originally posted by _d1
It's Habib Zargarpour who wrote about Perfect Storm.

Ahh yes thankyou for that! I didn't think it was Alex but I couldn't remember.

:scream:

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.