[MUD-Dev2] [DESIGN] Ray traced environments

Acius adamhelps at gmail.com
Thu Mar 29 13:04:08 CEST 2007


Tess Snider wrote:
> Look at this scene from Toy Story, using raytracing (and probably some
> material shaders):
> http://media.outnow.ch/Movies/Images/1995/ToyStory/dvd-film.ws/16.jpg
>
> Do we need raytracing to match that scene today?

Clearly not, as that scene was not ray-traced. Pixar films are rendered
using Pixar's PhotoRealistic Renderman, which is primarily a scanline
based rendering package. It does support some visual effects that use
ray-tracing, but in truth it's a hodge-podge of algorithms primarily
built on top of a scanline renderer.

To my knowledge, the only animation studio that is using ray-tracing to
render movies is BlueSky (Robots, Ice Age et al), and they built their
own. I watched a presentation on the making of Ice Age at SIGGraph, and
the presenter mentioned that they, uniquely, had decided to use ray
tracing instead of scanline rendering, mostly because it was what they
knew how to do; a good chunk of his presentation was showing how to get
ray tracing performance on par with scanline performance.

In the non-real-time world, ray-tracing isn't such a beloved technique.
On the one hand it's kind of slow, and you can fake a lot of its
benefits by doing something else (game studios aren't the only people
who want to get their render times down, albeit that movie studios think
in hours per frame instead of milliseconds). It's also been superceded
by higher quality methods in some cases -- radiosity is very difficult
to do with ray tracing, and I vaguely remember someone trying to propose
photon tracking as being even better than radiosity.

Sorry I didn't reply to this earlier, I wanted to double-check (I'm not
an expert on this stuff, I just pay attention in SIGGraph on occasion).
But "Toy Story" as an example of ray tracing struck me as kind of funny,
especially when given as a counterexample to scanline rendering; Toy
Story IS an example of scanline rendering.

-- Adam




More information about the mud-dev2-archive mailing list