“Transformers: Dark of the Moon” Breakdowns
Here’s another look at some of the breakdowns from ILM for “Transformers: Dark of the Moon”, including the Driller sequence.
You can find more of their amazing work here: ILM Visual FX on YouTube
Hope this keeps you warm while we cook up a few things and hope to have some more to coincide with new software releases.
Posted on February 27, 2012, in Example, unified sampling and tagged breakdowns, film, ILM, mental ray, Transformers, vfx. Bookmark the permalink. 7 Comments.
elemental ray 
cg is amazing,
would be interesting to know some more technical informations.
for example what type of shaders they used. all measured brdf’s? how about importance sampling? what type of indirect lighting? was it baked thru point clouds with renderman or mr only (if yes, fg or a custom solution)?
The Driller itself was mental ray. It performed the best for that sequence. I would *guess they didn’t use BSDF if they were wanting passes from the materials. It’s possible, but I don’t think the mechanism was built into 3.9. As for lighting, they tend to use a lot of their own setups.
own setups, thats the point i am after. 😉
its hard to get some information from ilm about their tech stuff. the physical rendering paper is a good resource but it does not contain some general information for their gi for example.
they use point based stuff in renderman, alot of prebakesteps. but it would be great to get some information how they lighted the driller shots for example with mr (or the help of renderman, dont know).
for the bsdf and framebuffer thing, it is a actual limitation of the mr core itself or is it a shader writing problem?
is it possible for you to tell us what was the render time per frame on this sequence and the hardware?
If I recall it was pretty high, dozens of hours a frame. But they render not only passes, but multiple layers (I know of a shot on 2012 with 80+ layers) and I think the times quoted were for a whole frame, layers and all. And I don’t know about their hardware. A few years ago it was surprisingly low-end but has probably changed by now.
They were not using Unified Sampling as we have described it in the blog. In fact we think they could have gotten better performance from brute force approaches on their materials.
I’ve worked several places where film frames are 10+ hours a frame for a single layer with different renderers (other layers may be faster) but typically speaking for mental ray jobs I have sub-3 hour frames for a usual 1080 HD frame.
Thanks David..that was really helpful..
Pingback: Recent Updates – Jonathan Moore's Journal