That’s certainly possible. But if folks on this site don’t know how ToyStory or JurassicPark were rendered back in the day, let alone like almost every movie with CG in it since, that would be a little surprising. But maybe?
That may be it. Renderman is of historical interest because it introduced shaders to the industry.[1] That's where the look of Pixar's early work comes from. Everybody else texture-mapped photos of the real world, or 2D paintings, while Pixar was generating infinitely fine detail with shader programs. This allowed the camera to get close to objects without the textures going blurry or blocky.
Today, shaders run in GPUs, and everybody does this, but in a different way.
Exactly this. The critically acclaimed and profitable films are in a sense just adverts for the tool. They're proof that you could not only have an Oscar-winning movie that used this tool for effects, you can make the entire movie with this tool.
See also: Disneyland. People really like Disneyland, and it's profitable for Disney, but it's also one enormous advertisement for their products.
Since Disney and Pixar merged their focus has been on making film franchises and the surrounding merchandise. So it's understandable that most people don't know that wasn't always their focus. When Jobs took over Pixar, he thought they were making a competitor to SGI. They had ambitions to sell not just Renderman licenses, but Renderman hardware. Jobs really wanted them to be a hardware company. They probably saw themselves as occupying the role NVIDIA fills today.
The feature film animation studio was sold to Jobs as a way to market their technology, in order to sell more licenses and create the market which would want their (still theoretical) dedicated GPU hardware.
Turns out the better play was making an animation studio that would rival and then eventually demolish and replace (via merger and integration) the iconic hand-drawn Disney animation.
I suspect the heads of the animation studio within Pixar knew what they were doing though.
Renderman has been used by pretty much every VFX studio for the last 30 years [1]. I remember downloading and dabbling with it along with BMRT [2] in the early 2000s. It was the first widely used tool to have a procedural first approach to rendering.
People forget that Pixar is a 3D rendering technology company first and only secondly a film studio. When most people out there hear Pixar, they think Toy Story and Finding Nemo, not even knowing about the technology they developed that makes those movies possible.
In the industry, studios will have different tech stacks that make them competitive for different areas. Some will have specialized simulations software for fluid dynamics, others really advanced rigging techniques that allow super realistic results faster than competing studios. A lot of it comes down to custom pipelines and integrations depending on their needs.
The off-the-shelf tools have come a long way for a lot of tasks, but studios still create really awesome in house software that in a bidding situation (VFX) can help a bit. Though to be frank the bidding process is an absolute mess.
When it comes to rendering, it all depends on the focus of the projects and the direction/needs of the studios. Though the end results may be similar, the way you get there will have varying degrees of optimization and flexibility depending what the stack is built on (besides the typical necessities of traced rendering).
The devil is really in the details. Pixar and Disney throw scenes at Renderman for overnight rendering jobs where each node loads up to terabytes of data per frame (geometry, textures, volumes...). So the renderer contains tons of tricky optimizations under the hood to support that crazy amount of data. Other renderers like Cycles use pretty much comparable physical models for everything and create almost identical output for simple scenes. But most of them would crash and burn spectacularly of they had to deal with scenes at the size these studios produce.
[Even though you could theoretically generate a lot of detail, especially textures, algorithmically on the fly, nobody goes for that: the only known good way to band limit textures for level of detail reduction without aliasing artifacts is to sample and filter them - generating the whole texture anyway.]
Training & tutorials for Pixar products used to be notoriously difficult (and expensive) to find, unless you worked at a studio that was paying millions for a site license of prman and engineers to integrate it with your studio pipeline.
It appears that Pixar is becoming a little more friendly to solo end users now that there are more rendering choices out there.
If you want to get into 3d art, I would definitely give Blender a look. Literally everything you need in one executable, albeit with a steep learning curve (but that's the case with most 3d software).
Blender has gotten considerably better. I think it will keep going strong due to the huge Unity crowd. Unreal shops are more likely able to afford commercial licenses on famous software.
I argue that Blender is right up there in terms of importance and significance in the OS world.
Please consider donating to Blender. I have pledged at least a low 6 digit figure over the past 20 years and I've loved seeing it grow. I don't get any returns obviously just glad I played a small role.
Not sure why people would downvote this. I guess supporting OS with donations is frowned upon here. Meanwhile everybody is more than happy to use free software here. I find this disgusting.
What I find remarkable about Blender is how much it's used by professionals for side projects. These are the types of people who would be very proficient in Maya and could use it for outside work (for basically nothing last I heard). Given how much the industry relies on open source it's only a matter of time until it gains wider use for big budget work.
Edit: I'm mainly talking about what I've seen in VFX here. No idea if this also applies to games.
From all the artists I talk to, Maya is still much preferred. That said, Blender is still respected. The reputation is that they're fairly close, unlike say, Photoshop and Gimp where many would not even give Gimp a chance.
Blender has been converging more towards industry standard UX practices, while GIMP sticks to its tried and true idiosyncratic way of doing things. It’s good GIMP exists, but it’s a failure in the sense of that it remains a programmers idea of what an UI should be. GIMP could have a had a lot more uptake if it wasn’t so stubborn. Something the folks at Blender understood at some point.
Management of Blender by Ton Roosendaal has been really good.
> but it’s a failure in the sense of that it remains a programmers idea of what an UI should be
Exactly, but that was also the problem with Blender for 15 years too.
Developers should have enough humility to let the UX design to designers and listen to them. Unfortunately, that's rarely the case according to my experience. The web sector is an anomaly in software development, because better UX translates directly into more revenue.
Maya is preferred because it's the industry standard in VFX. Studios have invested literally years and millions of dollars building their pipelines, and ripping out Maya in favor of Blender isn't going to happen, even if it saves a bunch of money on Autodesk licenses.
A friend of mine who spent the last 7 years as a Maya modeler (at ILM, DNEG, MPC, etc.) is now switching to Blender... for concept art. It's really taking off there, since concept is outside the production pipeline and can use pretty much whatever tools they want.
Production modelling is still overwhelmingly on Maya, mostly because of the pipeline, but also because Blender has some scaling issues with production-scale models.
Renderman 24 which is currently in beta and planned for release in 2021 will support Blender as a full authoring 'bridge' tool (same as Maya, Houdini and Katana).
"Literally everything you need" except for Boxcutter, Hardops, Atmosphere, and the other $200 in add-ons you need to buy to really be productive in Blender.
Maybe for a very specific workflow, but you don't need paid addons to be productive in general. I've put together quite a few massive projects quickly in Blender without spending a cent -- I'm sure those addons are extremely time saving for the specific use cases they're intended for, but vanilla Blender does have everything you need to get into (and even do professionally) 3D
That's fair. The problem is that Blender is very much a 'jack-of-all-trades' application. It tries to do everything from modeling to sculpting to texture painting to particle systems to rendering... that's something like 3-4 different commercial applications.
If someone is doing hard surface modelling in Blender, I would be shocked if they didn't have Hardops and Boxcutter (and MACHIN3Tools, but that's free). If they're doing environments or characters, maybe not (although auto-rig pro for character work is magic).
Totally agree. Blender does feel a little thinly-spread at times. In being ultra-flexible it's also not quite fully featured in many of its workflows. For someone like me, who just does generalist and VFX work, it's fantastic, but I could totally see where it might be lacking for a specialized application like hard-surface modeling.
Nah, I never used any of these addons although I know what they do, so they are far from essential. I use "Sverchock" for procedural modelling and "Animation Nodes" for procedural animation on the other hand both are free. Furthermore, the basic Blender distribution ships with a tons of free addons at first place.
Now should Blender improve their topology tools? Of course and I have no doubt with all the grants Blender foundation got that they will.
Khan Academy has a course called "Pixar in a Box". It's aimed at younger people (i.e. 10-15 years old), so the presentation is a bit cheesy, but I learned a lot from watching it.
Pricing is actually surprisingly cheap. I would have thought for sure one license would cost thousands, but it's only around $600. Out of the range of the average hobbyist obviously, but easily acquirable by any small business.
Also want to toss in that it is free for educational institutions as well. Just reach out to them and verify things and you can get a floating license setup with an agreed upon (iirc) number of licenses). For my university I got 100 licenses (includes 100 separate seats of Tractor, the render queue scheduler) with pretty much no effort, and it (RenderMan) was very easy to setup. Tractor is a different story...
Renting nodes makes sense these days? I've seen a mailing list post by I think Larry Gritz around 2000 noting that in case of serious productions, bandwidth for assets proportionate to the computational power required or provided would be infeasible for off-premises data centers.
It used to be a lot more expensive! IIRC, they had a per thread-ish licensing scheme for a while (maybe blocks of 4 or 8 threads?). Lots of software did in the late 2000s.
Then a few years ago, to stem the tide from Arnold (which was 1000 euros per node), they lowered their price to this current $600 price point and free for academic/non-commercial. They kept MPC/Technicolor and ILM as well as many smaller shops, but the other major studios either have their own renderer (including Disney Animation!) or use Arnold.
58 comments
[ 3.5 ms ] story [ 112 ms ] threadToday, shaders run in GPUs, and everybody does this, but in a different way.
[1] https://en.wikipedia.org/wiki/Reyes_rendering
See also: Disneyland. People really like Disneyland, and it's profitable for Disney, but it's also one enormous advertisement for their products.
The feature film animation studio was sold to Jobs as a way to market their technology, in order to sell more licenses and create the market which would want their (still theoretical) dedicated GPU hardware.
Turns out the better play was making an animation studio that would rival and then eventually demolish and replace (via merger and integration) the iconic hand-drawn Disney animation.
I suspect the heads of the animation studio within Pixar knew what they were doing though.
[1] https://renderman.pixar.com/movies
[2] https://en.wikipedia.org/wiki/Blue_Moon_Rendering_Tools
https://blog.siggraph.org/2019/01/leading-the-way-celebratin...
I thought renderman was Pixar’s in-house magic juice - the tech that puts them apart kinda thing.
The off-the-shelf tools have come a long way for a lot of tasks, but studios still create really awesome in house software that in a bidding situation (VFX) can help a bit. Though to be frank the bidding process is an absolute mess.
When it comes to rendering, it all depends on the focus of the projects and the direction/needs of the studios. Though the end results may be similar, the way you get there will have varying degrees of optimization and flexibility depending what the stack is built on (besides the typical necessities of traced rendering).
Some examples:
- Pixar: Renderman
- Weta Digital: Manuka [0]
- Framestore: Freak [1]
[0] https://www.wetafx.co.nz/research-and-tech/technology/manuka...
[1] https://web.archive.org/web/20210126083056/https://beforesan...
[Even though you could theoretically generate a lot of detail, especially textures, algorithmically on the fly, nobody goes for that: the only known good way to band limit textures for level of detail reduction without aliasing artifacts is to sample and filter them - generating the whole texture anyway.]
So I guess I expected last copy to be sold around Toy Story 1.
It appears that Pixar is becoming a little more friendly to solo end users now that there are more rendering choices out there.
I argue that Blender is right up there in terms of importance and significance in the OS world.
Please consider donating to Blender. I have pledged at least a low 6 digit figure over the past 20 years and I've loved seeing it grow. I don't get any returns obviously just glad I played a small role.
https://www.blender.org/foundation/donation-payment/
Edit: I'm mainly talking about what I've seen in VFX here. No idea if this also applies to games.
Management of Blender by Ton Roosendaal has been really good.
You would wish something like this for GIMP.
Exactly, but that was also the problem with Blender for 15 years too.
Developers should have enough humility to let the UX design to designers and listen to them. Unfortunately, that's rarely the case according to my experience. The web sector is an anomaly in software development, because better UX translates directly into more revenue.
Production modelling is still overwhelmingly on Maya, mostly because of the pipeline, but also because Blender has some scaling issues with production-scale models.
Happy to hear that Maya is still kicking strong.
If someone is doing hard surface modelling in Blender, I would be shocked if they didn't have Hardops and Boxcutter (and MACHIN3Tools, but that's free). If they're doing environments or characters, maybe not (although auto-rig pro for character work is magic).
Now should Blender improve their topology tools? Of course and I have no doubt with all the grants Blender foundation got that they will.
https://www.khanacademy.org/computing/pixar
You can also rent 10 nodes per week fot $50 (for example rendering on multiple nodes in a small render farm)
For non-commercial there is a free-non-commercial license so you can learn how to use it at no cost.
Then a few years ago, to stem the tide from Arnold (which was 1000 euros per node), they lowered their price to this current $600 price point and free for academic/non-commercial. They kept MPC/Technicolor and ILM as well as many smaller shops, but the other major studios either have their own renderer (including Disney Animation!) or use Arnold.