Cycles is hard to use for animation because it is very noisy, it's fast to get an initial image (great for tweaking) but it's slow to get a noise-free image. Also this includes pixar's "denoising as a post process" which is really amazing (and might even be applicable to cycles renders).
Cycles shading graph is very limited (but well chosen) compared to the flexibility of Renderman Shading Language.
Renderman's subpixel shading is just amazing, so getting crisp, antialiased renders even with fine detail is automatic, whereas cycles might require tweaking the settings or really cranking up the sample rate.
Cycles struggles and has memory limitations on large (even not so large) scenes.
Renderman produces very good motion blur without much performance hit, which is critical for integrating with real-world footage.
Renderman has a lot of features (similar to plugins) to integrate with external tech (e.g. you could plug in a custom hair generator step), which is critical for production.
Renderman has a lot of "scalability" features like caching indirect lighting into brickmaps, or loading large sets of tiled textures, that cycles (or even blender) just don't support.
Also, renderman has been used in production for decades, so a lot of places have built up large libraries of tricks, shaders or tools that work with renderman that they can use.
Renderman is well integrated with many render-farm technologies, so if you're working on a large animation there are some great option. Of course there are ways to do this with Cycles, too, but they are generally less mature and lack some production features.
With all of that said, cycles is still an amazingly powerful renderer, and completely free (not just free to use). You don't have to generate and install a license file to use Cycles, and if you want to tinker with the source code, you can do so! That's not going to happen with renderman anytime soon...
For a fair comparison we should compare PRMan's RIS with Cycles though, not REYES. RIS is what Pixar are focusing on and rendering their movies with now.
Yes, REYES is great at subpixel shading, rendering fast motion blur, using brickmaps with indirect light, and using very little memory for detailed displacement. But with RIS all those things are gone. PRMan's implementation might still be more efficient, I don't know, but with the switch to path tracing they are definitely giving up various advantages that REYES had.
Exactly, with the move to RIS, they're in pure path-tracing now, and they're still behind Arnold generally from a performance perspective.
Displacement and subd performance is still very impressive (compared to Arnold), but motion blur (especially deformation for curves and triangles) has a noticeable overhead now.
Cycles is pretty great. RenderMan's value is entirely in it's extremely fine level of control. It's typically used in production settings by RenderMan power users who can write custom shaders, complex interactions with non turn key systems, like custom simulators, and other production workflow tools.
It has been hammered on in many ways by many people for several decades and is focused on stable efficient production. For this reason it can take quite a lot of work to get great images out of it, the configurability and control come at a cost.
I think of light transport simulators as a completely different thing from production renderers. A simulator is like a camera that captures reality, where as production renderers are not bound by reality so they are more like fine art tools (brushes, rulers, etc.).
I dont know the current situation, but last time i checked cycles was sower relative to the competition and unarguably Renderman is far more battle-tested.
My experience with Cycles is that it generally looks almost great, except for a bunch of terrible looking fireflies (lone, overly bright pixels) for which there is no straightforward cure.
(I am very grateful for Blender and Cycles in general though).
Rendering, is a job in itself. There are people who do nothing but develop looks and write shaders all day everyday, and others who create and manage distributed rendering pipelines.
Cycles is great, but just like Mental Ray, in other applications, it's part of another package. Sure you can run them headless, but why? When there is something that is made for it that is so well documented and has great support solely for that purpose.
Rendering features aside, it's a great business and architecture choice.
It isn't even up for debate among anyone with in depth rendering knowledge. Cycles is free. That's the extent of its advantage. Renderman is faster, far more flexible, has true layered shaders, handles volumetrics, lots of geometry types, displacement, ptex textures, has better integrators, AOVs, a shading language, a detailed API, light path expressions... There are probably twice as many things I can't think of at the moment.
Two benefits were named:
1. Deciding to go with renderman gives allows you to tap a large talent pool that are already familiar with Renderman. Much larger than cycles.
2. Renderman works with a number of other 3d tools besides Blender. Cycles could work with other software, but I suspect this is a relatively niche group doing so. This means that other pieces of software could be used in conjunction with Blender.
Another way of putting it is that you can get a job working with Renderman.
What's really huge here is being able to interface open-source software with the commercial rendering pipeline that is used to make most of the movies we watch.
I had a big fight with my 3d modeling instructor who made us use 3ds max because he wrote the Adobe Press book on it, though he not very subtly suggested that we would have to pirate it in order to do our homework. I pushed really hard to be allowed to do work in Blender, wasn't a win at the time, but I think this is clearly the direction things will go - the days of paying thousands of dollars for content creating software that is primarily used by a handful of companies are numbered. Companies like Pixar and ILM should be contributing to projects like Blender instead of developing proprietary internal tools and marrying themselves to packages like Maya that aren't terribly better than Blender, if they even are better at all anymore.
Blender has some quirks, esp in modeling from what I remember, but it's solid, and more importantly, there are room for lots of tools, just not lots of tools that break the bank. This is a hobby for everyone it's not a profession for, and even the professionals are often contractors who struggle to maintain a toolkit.
> Companies like Pixar and ILM should be contributing to projects like Blender instead of developing proprietary internal tools and marrying themselves to packages like Maya that aren't terribly better than Blender, if they even are better at all anymore.
For anyone who might think the same thing:
1. This is not going to happen anytime soon.
2. The tools they use, proprietary or not, are better than blender. I know everyone loves to rally around open source software and Blender has benefited lots of people for a variety of reasons, but high end visual effects is not its place right now.
3. The large 3D programs are actually used by an enormous amount of people these days, not just the big VFX companies.
- Sony Pictures Imageworks open sourced Open Shading Language, Field3D, Alembic and OpenColorIO.
- Pixar opensourced OpenSubDiv
- Disney opensourced Ptex, BRDF Explorer, SeExpr and Partio
- Dreamworks open sourced OpenVDB
I wouldn't be too suprised if we would see an open source renderer within the next 10 years, especially if SPI is going to replace their in-house version of the Arnold renderer, as open source software has been extremely successful for SPI.
The open sourcing of C and C++ libraries in CG is awesome. But that is skewing the original point. Studios aren't going to replace their interfaces with blender and they aren't going to open source their own creation software. Most studios use a combination of maya, katana, 3ds, houdini and vray.
What you listed are mostly formats that are beneficial to everyone to share (and don't forget OpenExr). OpenSubdiv lets others match renderman's subdivision. Also Imageworks' open source push can largely be traced back to Rob Bredow I believe.
Autodesk recently made their software (including 3ds Max and Maya) free for educational/non-commercial use for 3 years. No watermarks or other restrictions.
I'm not sure why you think circumventing the license poorly is any better than using cracked software. I won't call it theft (I don't think illegal copying should be considered theft) -- but I also don't see any meaningful distinction between how you break the license and use software you haven't paid for.
So far I must say cycles is pretty great