Best renderer for animation?

Hello there!

I’m planning to work on a big animation project, and am still wondering what render engine to use.
I’m not really familiar with render engines, except for cycles and BI. But I’ve used Luxrender before, and only that.

So, if you please, would you mind telling me a good render engine for an animation project. And telling me the downsides and upsides for each one? :wink:

If it makes a difference, it’ll be a car animation. I’m planning to make it as realistic as possible without VERY high rendering times.

Thank you very much =)
You help is much appreciated!

Thank you very much, your links helped me, I’ll read em later on.

I’ve taken a look at Indigo, it seems to be good. I’m not sure about its render times though…

Sorry for the short reply, I was in a hurry and out of town for 2 days - anyways:

For animations, where I can use Direct Lighting only (fake GI and no caustics) I´d go with (random oder):

Free ones:

  1. Blender Internal - awesome control over everything but a bit elderly by now :wink:
  2. Cycles DL Kernel - Fast with CUDA, still lacks too many features though for me to use it in a full production.
  3. Luxrender - Luxrender is often underestimated for animations, but it´s DL kernel and termination settings for rays are quite powerful and Luxrender is a very mature render IMO.
  4. Yafaray - the FOSS verison of vray. A kickass biased renderer, which still lacks some features but it´s being worked on.

Commercial ones:

  1. vray - great renderer, still superior to Yafaray (renderpasses, SSS and more), comes with 10! node licenses for networkrendering and quite cheap as standalone with a great Blender exporter.
  2. Thea - very good blender integration, very good material studio and lots of render kernels, costs more than vray though.

You see those are the usual suspects, however:
All but Blender Internal and Cycles have the disadvantage of having to export each frame of any animation to render it.

Which is also why I don´t name Octane here. It´s DL kernel is incredibly fast, node materials and the library are awesome, also the other kernels, but I only use it for turntable animations for product presentation/vizualisation(doable from within octane) or stills, because exporting the scene, importing to octane and voxelizing often takes more time than the render itself and has to be done for each frame. And on complex animated scenes with lots of textures your video memory soon can limit the whole show (Cycles suffers the same fate)

Indigo and Indigo RT are both great renderers, but maybe not fit for animations. It´s great for archviz, both interior an exterior and does a great job with complicated lighting equations. RT is cheap, but has no lightlayers and SSS and something else I forgot. I got a license, but in all honesty barely use it.
It´s also not too fast, not too slow either. OpenCL can only be used with pathtracing, but not with Indigos strenght, the bidirectional pathtracing.

If you need/want GI and caustics, maybe dispersion, you can eliminate BI from aboves list, all others are still a valid choise.
Vray and Yafaray are both biased and a good choise for animations with believable results in a decent time.
The rest is unbiased and depending on the kernel you use it can be a while until you render a noise free image.

As you see in the list in one of my links there are a good 100 renderers out there, from free to 10.000Euro+ per seat license, all with different strenghts and weaknesses.
It´s hard to generalize and make a recommendation. To choose the right raytracer you’ll have to have a basic understanding of raytracing and the methods, know their pros and cons, weigh it against your requirements and look at how fast you can do what for what cost and top of that it should work within your pipeline.

And the choise of the “best” raytracer is an esotheric/religious matter or personal preference really. Where one artist strifes for a 100% physical correct representation of the real world even if it takes a year for a frame, the other one wants to fake as much as possible it the least time possible to deliver a believable result.

One gig might requires you to render a product visualization of a diamond ring with dispersion and accurate caustics in the gem and from the metal, while the next gig requires you to render a fast moving car as realistic as possible moving in a scene in 2 days, where the viewer sees the car for 2 seconds rushing through the screen and no one cares if the reflection in the paint are 10cm off or the caustics from the rims are missing - the viewer will much more notice if the car has odd proportions or is moving oddly, so better spend more time on that.

http://www.luxrender.net/wiki/SLG

Never used it but heard good things

Posts like this make me curious.
So what´re the good things you’ve heard?

The best rendering engine for animation is Renderman. Or a Renderman-compliant one that is in your affordable price range.

At the moment the factors that affect your decision include

A) Budget > Open source, several hundred dollar software, or top of the line renderers that are $1000 or more.
B) Available Hardware > CPUs available, GPUs (Cuda/OpenCL), RenderFarm, Online Renderfarm Service
C) Raytracing vs Scanline Render and Reyes Rendering (Renderman) > Cars benefit from Raytracing, but scanline rendering uses cheats to look realistic, rather than doing the brute force calculations of raytracing. Raytracing allows for more realistic reflections and shadows, as well as glass and caustics.
d) amount of time to learn the software > programs like LuxRender are easy to learn, because you are simulating reality, while programs like renderman require a lot more technical knowledge. On a scale BI is closer to renderman, and takes more work to look good, while Cycles is closer to Luxrender, and takes less work to get photorealistic renders.
e) Are you going to use layers to composite the final scene at the end, or will you render as is. Luxrender does not allow layers, although it does allow light groups. Cycles does not have layers yet, but it is on it’s way.

A review of Open Source software that I am familiar with.

Cycles is incomplete at the moment, so there are a lot of features missing, but It can render cars and metal ok already. Layers are coming soon so you will be able to render images as separate passes and composite them. It is Raytracing, Montecarlo method, and attempts to be along the lines of Arnold Renderer. It is CUDA enabled or CPU only at the moment. It only does path tracing, so it takes a long time to render scenes lit with bounce light. If you have the personell, time, and resources for renderman, it is a great choice, but from your first post, it seems like Renderman might be a stretch for you in this next production.

LuxRender attempts to be completely unbiased, so You cannot render layers and you can’t do any tricks with it. It uses brute force raytracing, and has a wide selection of realistic materials to choose from. It has OpenCL GPU acceleration, or CPU. You can use light groups, which allow you to relight the scene after the render, but you can’t split diffuse, ambient or specular passes.

BI is a scanline renderer. It is old, but relatively full featured. It’s not realistic by default, so it takes a lot of work to get good results (relative to cycles). By default BI is faster, but as the quality improves it ends up taking longer.If you are really doing Car animations, this is probably not the choice for you, but BI makes an excellent tool to supplement many of cycles shortcomings by creating additional layers.

Yafaray might be a good choice. I have not used it, but It leans closer to Cycles and Luxrender. Some say it is faster, and therefore might be better for animation than luxrender, but it is not as realistic out of the box.

Renderman is a Specification created by pixar not a software. Pixar sells Prman as a renderer that does the Renderman spec. It is $2000.00 (price may be wrong). But since it is a spec, there is also 3Delight, where the first two cores are free. But the open source versions are Aqsis and Pixie. Prman is used in most of the major blockbusters that come out today, and the spec allows you to get similar results from any Renderman program. There are two catches though. First Renderman is complicated and takes a lot of training to learn (Some say years). Second, Although Prman has a new Raytracer that is fast, your mileage will vary with the affordable renderers when it comes to raytracing.

I hope this helps.

Wow, thanks for your replies!
I’m really glad I started this thread, wouldn’t get as much help anywhere! =)

In all honesty, I’d love to use BI, because I’m really familiar with it, easy to use, pretty much full-featured (or maybe almost)… and is fast. But, as we all know, it’s hard to get good results.

However, take a look at this

That’s a pretty neat result :stuck_out_tongue:
I’d love to have my cars look like that.

Also, I’m going to do some research about Yafaray
I’ve never liked Luxrender to be honest, and cycles is great but it’s too slow for me.

Thanks a lot for your replies! :wink:

Well in reply to arexma, it’s free, it’s supposed to be wicked fast (exceptionally so), and it has the same realism as lux.

Hmm, SmallLuxGPU, I’ve heard of it, but never tried it.

I found this one Youtube, it’s rendered with SmallLuxGPU


(the actual animation is the first 30 secs… the rest is just repetitions :P)

Looks rather nice tbh…
Thanks, GraphiX =)

Edit:
Here’s this one too, rendered with SmallLuxGPU

And this one’s rendered with BI

SLG is no Renderer, or not really, it´s a proof of concept, a sandbox for Luxrender.
It´s more targeted towards techsavy people and developers rather than endusers.

Everything that works, is stable enough or feasable to implement, is taken from SLG and gets implemented as feature into Luxrender. And the pure GPU raytracing is implemented into Luxrender as Luxray and is somewhat experimental at the moment.

However, learning Lux is a good thing, no matter how good or bad their GPGPU support currently might be, it works with many applications, does all from biased direct lighting over photon mapping to full GI MLT pathtracing, has a good material system that supports volumes is constantly under development and their idea to keep it hardware independent and take the best out of CPU+GPU+network is just grand.

Its kind of difficult to answer I feel, there is so many render engines to choose.
Today I primarily use Vray and 3dsMax, since I mostly working with still renders. I have rendered some animations with Vray too, but its kind of a difficult process I feel.
You have pre-render your calculations and then render your final render. I believe if you have render farm it shouldn’t be that hard. Only thing is that you have to render out all those passes…

I haven’t tried so much more in a pipeline for work, but at home I tried Mental Ray and Brazil. Both engines are a lot more friendly on animation rendering. But Mental Ray is slower than Vray when it comes to GI.

animation institute in Chandigarh