DreamWorks' OpenMoonRay Renderer Code Published

[Frank Cox]

https://www.phoronix.com/news/OpenMoonRay-Published

Last summer DreamWorks announced plans to open-source MoonRay, their production renderer used for films like The Bad Guys, Puss in Boots: The Last Wish, and other animated films. Today they have delivered on that exciting milestone with publishing the open-source code.

Last month we heard from the Academy Software Foundation's Open-Source Forum that the OpenMoonRay release was very close and it turned out to be the case: OpenMoonRay is now available.

OpenMoonRay is available via DreamWorks Animation's GitHub. This professional-grade renderer is available under an Apache 2.0 license.

MoonRay is DreamWorks’ open-source, award-winning, state-of-the-art production MCRT renderer, which has been used on the following feature films:

How to Train Your Dragon: The Hidden World
Abominable
The Croods: A New Age
Trolls World Tour
The Bad Guys
The Boss Baby: Family Business
The Bad Guys
Puss In Boots: The Last Wish

MoonRay was developed at DreamWorks and is in continuous active development and includes an extensive library of production-tested, physically based materials, a USD Hydra render delegate, multi-machine and cloud rendering via the Arras distributed computation framework.

https://openmoonray.org/

[Jon Dent]

As evidenced from the list of titles above, it's a very nice renderer. And It is the only major production renderer that is (now) open source.

Still, with the rendering market as crowded as it is, I wonder who's really going to care...

[Marcel Birgelen]

Yeah, it all sounds pretty noble, but I guess next step will be firing their software development team?

Maybe a major open source project like Blender could at least cherry-pick the stuff that makes sense for them, otherwise I'm afraid it will be abandoned pretty quickly.

[Mike Blakesley]

This whole topic is basically Greek to me but I'm curious, why would they do this? What is in it for them?

[Frank Cox]

One reason is to share the cost of maintaining the program. Dreamworks isn't in the software business; they're in the movie making business and software like this is one of the many tools that they use to make movies with. It's part of the plumbing, really, in the same category as things like microphones and light stands.

If they can offload some or all of the ongoing upkeep and enhancement of the program, they're not actually losing anything and could be gaining a significant advantage in terms of not only the costs involved in the program itself but also the possibility that some "outsider" will add something to the program that will cause their animators to say "Eureka!" and make the next number-one-on-the-chart animated movie.

The upfront additional cost to them for open-sourcing this program: zero.

The possible future advantages: Currently unquantifiable but ultimately greater than zero.

[Marcel Birgelen]

This whole topic is basically Greek to me but I'm curious, why would they do this? What is in it for them?

The software we're talking about is a so called 3D renderer. It's the part of the 3D animation software that translates the wireframe graphics into 3D images. It calculates the light shading for each and every object in the scene, calculates shadows, applies transparancy, reflections and so on. The process is a very computionally intensive one and is often distributed over large so called rendering clusters.

Most effect houses and animation companies use "commercially available" software for their rendering tasks, it's not common for them to roll their own, as maintaining such a piece of software is a pretty expensive business. Pixar, for example, has their in-house renderer called RenderMan, which they also license to third parties. But they were about the first one in the game, so they had to create their tools from scratch. RenderMan has been commercially available for at least two decades now and has established itself as an important piece of software at many animation and effect houses. MoonRay hasn't seen any such proliferation, unfortunately.

The problem I see with something like MoonRay hitting the Open Source world is that it may have made a big difference when it did so, like 15 years ago, but a lot of those open source renderers nowadays are so advanced, the question is if there is something in MoonRay that can't be found in one of the advanced open source rendering engines.

Maintaining such software requires a lot of very special skills. Back when I still was a university student, we had a computer visualization department that did some stuff with real-time 3D visualisation, non-real-time raytracing, etc. Stuff that was still comprehensible and maintainable by someone with some (by modern standards) basic knowledge in the field. Compared to current technologies, those things were crude toys, there is no way I, for example, could be remotely useful in maintaining a modern OS renderer like LuxCoreRenderer.

[Jon Dent]

The other aspect is that render engines usually have to choose between being fast and being capable of handling massive amounts of data. Engines like LuxRender, Redshift, or Cycles emphasize speed and are geared towards smaller operations that have limited computing power. Engines like Renderman, Arnold, Mantra and MoonRay are meant for large studios, and they can handle huge amounts of geometry, textures and volumes, but they're not fast, which is why big studios have huge render farms to support them.

Now there is some overlap, as ILM has used Redshift and freelancers use Renderman, Arnold and Mantra, but they are the exception.

That's why I wonder if MoonRay will get any traction out there. The big studios are all covered, and it probably doesn't hold a lot of appeal for freelancers who will have to develop completely new tools to integrate it into software like Blender. Not to mention that it probably isn't going to be the fastest car out there, and speed really matters to the time-is-money smaller studios. It also may have some unusual quirks given that it was initially designed for in house use by Dreamworks.

Incidentally, in house render engines are becoming more common than they used to be. In the last decade Disney has launched Hyperion, Weta Digital has Manuka, Illumination has their own renderer, and Dreamworks created MoonRay (and this engine superseded whatever they used for the Shrek movies, which was also an in house engine). Each one is unique and build to handle the very specific issues and workflow of it's studio.

[Marcel Birgelen]

We're going a bit overboard here, but it's a film forum after all... so what the heck...

From what I've heard, Hyperion is essentially dead now and all of Disney is forced to move back to RenderMan or other commercial renderers, the latest job cutbacks also aren't helping, apparently... It's interesting to see how there can be so much internal competition between what's essentially the same holding corporation.

Hyperion, was apparently created because RenderMan REYES was too slow for their operations, they wanted, at least intermediate results, to be available faster. So, instead of working with the RenderMan team to build a faster RenderMan, they built Hyperion. Afterwards, Pixar built a faster rendering mode called "RIS" and that's what they're apparently are all switching now to... Big corporations have a tendency to waste resources on a whole slate of expensive parallel projects...

Regarding Manuka, I was under the impression that it died somewhere back in 2015-ish, as I've not seen anything new from that front. I've been a long-time user of NewTek's LightWave, by far not the most powerfull 3D software out there, but it has had a long history and has been used in many TV shows and movies over the last decades, but as of last year it's clear it's now dead. The reason I've kept it around all those years is because I'm familiar with its workflows, switching to Maya, Blender or something else will take me months to get anywhere close to the speed I was before. It's clear that it's almost impossible nowadays to keep up with the top dogs. You see the same thing happening with game engines. While most big game studios have maintained their own 3D engines over the years, you'll see that many projects will eventually be shifted over to the likes of Unreal or Unity, simply because it's impossible to keep financing such complex technology in-house.

[Jon Dent]

Disney was using Hyperion as of Strange World and it is also the renderer for Wish, which releases later this year (according to the blog of one of Hyperion's developers). Unless them switching back to Renderman (which would be a colossal realignment) is a very recent thing it doesn't appear to be accurate.

As far as Manuka, they were still releasing papers right before COVID hit, and it was used extensively on Guardians of the Galaxy 2. It's hard to find information on Avatar 2, but Manuka is mentioned in a few articles about the technology of the movie so it seems like it's still alive and well.

I've never used Lightwave but it truly seems to be the software that will never die, and it was THE 3D software for television for a long time. It reminds me a lot of Pro Tools: maybe not the newest, maybe not the fastest, but after being top dog for so long a lot of artists are comfortable with it and know how to use it.

Game wise I agree that things are consolidating around UE4/5 and Unity. There are still some other engines out there (Decima, IDTech, Frostbite, whatever jank Bethesda uses, etc...) but they're either not for sale, or too difficult to work with.

And 3D rendering was (and still sort of is) one of my hobbies, so I definitely have too many opinions on it not to spew them out

[William Kucharski]

Which may or may not be part of the problem.

The biggest complaint I've heard about Strange World was the generic nature of the animation. Many had said that you could have cut and pasted characters between it and Lightyear and never would have known they were from two different films, where in the past Pixar and Disney at least tried to give each film a unique look.

Though Disney films were a bit more generic, Pixar certainly tried to use each new feature as a technological showcase; perhaps not any more with modern economics and the list of flops both they and Disney have had of late.

[Bobby Henderson]

I'll look at this stuff from time to time to see if it will have any impact on general purpose graphics applications. I suppose companies like NVidia could be interested in things they could cherry pick from Moon Ray and incorporate into some of their own in-house stuff. Adobe has been adding more pseudo 3D bells and whistles to applications like Illustrator. I guess that's a consequence from Adobe buying Allegorithmic, the company that makes the various Substance 3D products. I have no idea if or when Adobe will add a full-blown 3D modeling application to its arsenal. The Substance products are in a separate subscription tier of Creative Cloud. Those use NVidia's Iray renderer. I think the 3D stuff in graphics applications like Illustrator and Photoshop use Adobe's Mercury renderer.

If 3D printing methods really take off in the sign industry I might get more serious about doing 3D modeling work. Two critical issues are size limits and product durability. Outdoor signs can get really big and they need to last for many years. Our weather gets pretty brutal here; aside from the severe storm threat in the Spring and early Fall the wind just blows like crazy all the time. A sign with any sorts of fragile parts can fall apart in a short amount of time.