Just had a sort of random question and wanted to get your opinions. So I am looking for a good quality render for Maya. I am using maya 2016 and do have mental ray, as well as renderman. I know there are tons of renderers out there and wanted to get everyone’s opinions on what they consider to be the best one. I am new to modeling characters and have primarily done environment and prop modeling, other wise I will usually render out effects. I want to find a renderer that is best suited for movie quality CG (Pixar movies, Final Fantasy, etc.) and photo real renders. I have had a lot of success with Mental Ray in the past, but would like to branch out a bit and see what else is out there.
There’s no “best renderer” they have their strengths and it depends on what you’re doing with it.
My recommendation is to not get too hung up in plugins, Mental Ray is pretty solid and will do what you need to, and it’s already included in the software.
Just go with one and do your best.
The basics transfer between applications.
To me Maxwell is best and makes the most sense but all are very good.
If you wish most apps have demos.Try them and maybe one will impress you.
Good luck.
Personally I’d advocate for Vray. I’ve used Vray, Arnold, Renderman, Mental Ray, and Maxwell Render for finishing work. I’ve also used Modo’s renderer in comparison.
Vray is the one that has the least amount of fuss from test to final quality as far as overall settings to configure overall. Plus it also has a basic light/shader/object level approach for troubleshooting detail and noise based on your needs.
Renderman, Arnold, and Mental Ray often have a top down approach that balloons out rendertimes. With Vray you have per light, shader, and item, level specific features. Imagine giving 5 motion blur samples to only the wheels of a car to get the proper blur in sub frame times whilst still giving the rest of the scene 2-3 samples motion blur. Makes a massive amount of difference in render time having that kind of control.
Often people don’t learn all the renderers to this level though. I’ve seen arnold and Vray shaders setup up at default and the engine samples are then handled to brute force everything at the AA engine side instead of the shader subdiv side.
You can’t deny Mental Rays hyper tweaker world of different rendering looks and ARnold’s overall fast feedback approach but I feel Arnold and Mental ray sputter out during the final quality phase in comparison to Vray’s very slightly slower setup time to a very clean and crisp speedy final render.
Thanks for the info! I’ve always wanted to mess around with VRay. In school I remember seeing it in action in 3D Max (before it was available in Maya), but never really understood it. I’m going to check out renderers and see what works best for what I want to achieve.
(This isn’t an “I’m just learning Maya, what should I use” question, so I don’t think “don’t worry about it” is the right answer.)
“movie quality CG (Pixar movies, Final Fantasy, etc.) and photo real renders” is quite a range–that includes just about everything, since Pixar movies are the opposite of photorealistic.
I haven’t yet seen a really good comparison of major renderers. There are some pages going “I tried to make these look the same in multiple renderers, here’s what they look like”, which isn’t very interesting (that seems like it’s just comparing the author’s skill level in the renderers). It would be nice to see some in-depth workflow comparisons, eg. showing the practical experience of common tasks like creating SSS materials for a character. It would only be useful if it was by someone with experience in multiple renderers (a Vray expert showing how much he stumbles in something other than Vray would be fairly uninteresting)…
There’s a real time investment in learning a renderer to the point where you can decide if it’s actually useful for you to learn, and the lock-in is severe (recreating materials for a scene because you changed renderers), so content like that would be extremely useful.
In most cases if you are working hard with one renderer you will discover some problems and see its limits. If you do not aware of any problems there is no need to switch or learn another one.
We used mentalray a lot and had to fight with its limits. Especially rendering volumes and particles are ot the strengths of mr. The we looked for other ones. But if you really need a production renderer, what means it has to render particles, volumes, hair, motionblur, depth of field, is able to do procedural geometry, support archives and instances, then the list of renderers becomes quite small, five or six renderers, that’s all.
Very interesting post Casey. I have the most experience with V-ray myself, trying out Renderman at the moment to see what it’s about, and V-ray makes the most sense. V-ray property/displacement nodes beat the hell out of Renderman’s share nodes, much much easier to use on a very large margin.
What is your favourite AA? My studio uses Lancoz on every thing and our senior is very strongly against using anything else.
The problem is, in a real world production one usually doesn’t have the time to tweak every single light, shader and whatnot.
This is why mental ray took a quite different route in the last few years.
In 2016 there are only 4 “Quality” slider: Overall, Lighting, GI, Material.
That’s all. And this is a good thing, because at the end of the day it doesn’t really matter if the rendering took 10 or 15 min. But if you saved a few minutes of YOUR time because you didn’t need to adjust hundreds of sample knobs than that’s worth the extra rendertime…
Really??? Lancoz is the worst! It sharpens your image and introduces artefacts that are nearly impossible to solve. One should only sharpen the rendering in post. (Just as with an DSLR RAW Image.) These are slightly blurry as well, so that one is able to make it more crisp afterwards…
That’s my impression as well Stefen. My senior claimed Lancoz creates better Zdepth edges when apply to objects. I personally use triangle a lot based on my searching and reading quite some years ago. But I’m not an expert when it comes to comp so I really don’t know. If you can shed me some lights I would appreciate.
I’ve been in “real world” production for twenty years.
Try working at a high end facility next to another user of the same render engine who can go to the supervisor and say, "yeah I’ve taken 20 minutes off the render time. They now render at one hour twelve minutes on average instead. OH how did I do that? Well the lights were at default 8 subdivs, the DMC was set to… the shaders weren’t tuned properly for the glossiness setting used and I was able to then bring down the overall ADMC etc etc and just let the AA clean it up slightly. Motion blur … DoF… passes etc etc.
Once you have that person there to trump a top down engine setting you will undoubtedly be taken off setting up shots.
Arnold and Mental Ray and all of them can be hypertweaked like this. I didn’t mean to say Vray was the only one but that Vray does overall have the easiest set of parameters to learn overall to basically get rid of either shader glossiness/shadow/blurry value noises OR the GI blotchiness or noise.
Arnold being the closest to Vray. The one thing I see with Renderman as of RIS is that it can handle quite a bit of data. There are cases in which just being able to handle data is often trumping rendertimes on large productions. A veritable CAN we render this without crashing vs anything to do with render time. Mental Ray has quite a few different solutions. More so than the others because many different methods were invented as it was developed to solve certain problems. This is why mental ray has quite a few different ways to do things. Even the amount of GI solve methods is quite interesting as the tech evolved.
Vray had more methods in the past than it does now. Chaos has simplified Vray as newer methods become easier and more relevant.
Arnold was chosen for Jack the Giant Killer at Digital Domain and was barely ready for prime time back then. It has come a long way since. Meanwhile Vray had been used for Tron and almost everything since. With exceptions being Renderman for Transformers as legacy development was used as sequels were made.
Vray had proven over and over to be the fastest method from A to Z on so many jobs at Digital Domain and this was with world class talent that used all the engines and were able to make, forced to make, objective comparisons for budget reasons on films that spent hundreds of millions of dollars and wanted to have the best solution.
There were cases when Mental Ray’s SSS was used instead of Vray because of the custom shaders developed on previous productions like Benjamin Buttons but overall as Vray was updated DD we were able to use Vray’s SSS on Xman First class and other things moving forward.
I can safely say that Arnold is nothing to scoff at vs Vray though. There are some major improvements since the Jack days. Arnold overall has a bit more of a complicated base shader than Vray as far as attributes to tune giving the artist a bit more tweaking ability.
I can offer my opinion here and I can easily accept critique that it may be biased towards Vray and can leave it at that but I will say that my opinion is built on being around top tier TD and Lookdev people from around the planet on many commercial and films during my tenure at Digital Domain so my opinion stands on their shoulders as well.
NOW…
AA filter… Best balance between soft gaussian and Lancoz?
Triangle. You’ll get sharp edges without the over sharpening and you wont have cheated the AA with a soft look like gaussian even though gaussian does nice work of motion blur.
Do a Gaussian, Triangle, and Lancoz render of a car or zbrush creature sculpt.
I like triangle filter size to 2. I also use Triangle in Renderman as well. Its the unsung clean filter.
Interesting post. I have almost no experience with Vray, only a bit mentalray and Arnold. To me Arnold seems to have really only a few options where I can tweak the renderer. And compared to Vray or mentalray the implementation of shaders or procedurals is really easy. Do you think Arnold is more complicated to use than Vray?
vray would probably be my choice as the most well-rounded renderer right now. I’m not a fan of how often they’ve raised their prices over the last 3 years, but that’s life. It has a long history of ready solutions and workarounds, is fast, and is quick to update its UI and workflow.
renderman would be the next major one I'd consider, but most of its tech is new, so I'm sure as more people use it with its newer raytrace method, we'll start seeing where it has problems. Of course you get their legacy renderer as well if you want to dive into all that.
arnold is nice, but IMO too expensive - especially scaling up on a render farm. I've never seen anything other than realistic renders from it. Has anyone ever seen it render toon-shading or line art or anything NPR? From what I've heard it's render times for finals are a little on the higher side - so it's mainly aimed at large studios with huge render farms that can handle the longer render times.
maxwell is nice and accurate, but really slow. Probably can't do NPR either.
The others I don't think are developed enough to want to use them as a primary renderer, but look promising for someday in the future - like corona, etc
Mental Ray's MILA shader is pretty awesome, though people who don't like nodes or the concept of layering won't like it. They seem to be aiming for offloading a lot of render calculations onto the GPU which sounds like a pain in terms of setting up a render farm to have [u]both[/u] strong CPU and GPU. I like that I can always find a workable solution given how many things have been crammed into the engine over the years. The problem though is how often you need to look for a solution vs things just working like they would in most other engines.
I also never use Lancoz filtering. I’d even go so far as to say any sort of raw-data pass like zdepth, I’d think you’d want to stay as far away from lancoz as can be.
No, I don’t think Arnold is a more complex renderer than Vray at the UI level. I mean, not initially. Arnold has a quality slider the same as the newer mental ray goes.
What I ultimately mean is that there are things about Renderman, Mental Ray, and Arnold where I feel I can’t quite get at everything I’d like as easily as far as engine performance without getting into command line setups.
Vray also has more power as a standalone in a command line environment like Linux where you can make it do things that aren’t out of the box. I have a quick render tool that uses Vray standalone to fire off a quick render based on some very very low quality settings. This sends what I am looking at to the renderer without changing the render settings in the Vray settings UI.
Where I found some trouble with Renderman RIS or Arnold is when I want to achieve very specific results in regards to breaking up the engine performance on a per item or per feature basis I didn’t feel like I could get under the hood enough.
In Renderman I setup quite a few scenes to test over last year.
There is a section in that thread where I am testing geo light in a lightbulb that should cast caustics onto a floor through the glass. Or also a scene with a car lit only by indirect lighting, meaning all light hitting the car is lit by the bounce only. This is a great stress test for GI speed and how far you’d have to go to get a clean GI in even a dire situation.
There is a point where I notice that the RIS render could render the car wicked fast without GI. It practically acted like OLD Reyes. But once I fired up GI it went on forever. The only parameter I could get to was how many times I wanted bounce and Renderman just tries to keep cleaning it up. Hours and hours over night can go by. You see these breakdowns of the Good Dinosaur which look fairly generic overall with cartoon quiality characters the like Pixar is famous for. You see these quotes where the frames are taking like 40+ hours.
Its quite rare to have a Vray render ever ever ever go that long on a modern dual cpu setup.
Not being able to control the GI much in RIS besides just if its on and how many “indirect” bounces are used makes me understand that Renderman is meant for CAN it render it AT ALL kinds of large data projects where large render times are not an issue vs just being able to achieve the image.
Where as in Vray you can pick between three different GI solutions AND you can pick what is used for the primary and secondary individually. AND you can learn how each on works and how each method has its own flickering problem solves. The irradiance prepass is a really great solution for even low quality GI solves. It gives you multple ways to fix a problem based on the situation.
Mental Ray at least added the option to solve the GI on a GPU which then does the rest CPU. I suggested this in the Renderman forum. This idea would also make Vray quite powerful. Stay tuned.
Point being that there are ways in Which Vray lets you get under the hood the way Mental Ray does but it has refined that methodology and has rebuilt the UI over and over so that most causual users can just get renders out. Same with arnold.
Yet with Vray you can tweak individual GI methods, not to mention almost everything else, where as Arnold and Renderman are basically… ok… GI is ON… and its this many bounces and… you have to set the quality high enough to get rid of the problem and whatever the rendertime is… it is.
Vray lets you split hairs in every category. This is why it has remained near the top at DD and many other facilities. BUt…
If you aren’t a hypertweaker that has to produce minute per dollar results and you can just throw the render farm at the big Smaug shots as long as it gets finished then Renderman or Arnold fits the bill because the initial setup is quick.
The feedback loop is quick and, like mentioned above, you can just assemble and use the time rendering instead of setting up. But, in small commercial houses where a week can make or break a budget Vray allows the best quality to performance controls I’ve used yet.
I will say that Renderman Reyes had the old brick map method for GI but that quickly could get bogged down once you had to do something like grass with GI. So bogged down that you wouldn’t even see your render start until 45 minutes into the initialize render so the GI calculation could get sorted. This is why RIS is like magic in one sense. You can see the render really fast.
RIS is so new it is hard to critique. Because Vray for maya has been refined over 5 years now. RIS may end up there too. Arnold has come a long way too. It went from being something I liked due to the preview but wouldn’t be caught using it for finals to something that has taken a fair amount of market share in Hollywood films.
Ok sorry about the biblical testimonial.
I honestly have a ton of fun using Renderman. I do like the SSS shader in Rman quite a bit. None the less. Yet overall I always end up back in Vray to get things finished.
Wow, I did not think my post would start such a great conversation on the topic. Like I said in my initial post I have only rendered really with Mental Ray, but after hearing what everyone’s thoughts are on the different renderers I am definitely going to be trying some out and seeing what I can get out of them. Does anyone know if the demo versions of the renderers like VRay have like a watermark or anything like that on the final renders? Or is it just a full usage for a limited time demo?
