1
$\begingroup$

What are some of the challenges of Offline Rendering? I know one problem that is faced is performance and rendering times, are there any more issues?

Path Traced photos seem to have reached the point where it is hard to distinguish a CGI image in comparison to a real photo so being able to render a photo-realistic image doesn't really seem like a challenge anymore.

Is offline rendering just about refining current techniques to find an improvement in render times now and not so much about photo realism?

(Grand Moff Tarkin in Star Wars: Rogue One was entirely created through CGI since the actual actor had died)

|improve this question
$\endgroup$
3
$\begingroup$

From the presentations on non real time rendering I've seen at SIGGRAPH, the answer is basically yes, a lot of time is spent making render times faster. For instance, even with huge amounts of ram, the amount of data needed to render a scene is larger than can fit in memory all at once, so strategies are developed to minimized disk access during the render process since it is so slow.

However, I have also seen a lot of effort and research put into simulations - such as realistic foamy ocean tidal wave simulations.

I guess once the rendering is realistic, the next things to work on is the simulation that is being rendered.

|improve this answer
$\endgroup$
  • $\begingroup$ When you say simulation, do you mean simulating the physics of things like hair or rendering things more accurately? $\endgroup$ – Arjan Singh Jan 14 '17 at 5:41
  • $\begingroup$ I mean things like simulating the motion of hair and water and things yeah, not the light physics. $\endgroup$ – Alan Wolfe Jan 14 '17 at 5:43
  • 1
    $\begingroup$ ... although light physics is sometimes a part of the more advanced simulation. For instance, in the movie interstellar, they rendered the black hole based on what we know of black hole physics and actually ended up publishing a research paper about it! $\endgroup$ – Alan Wolfe Jan 14 '17 at 18:22
  • $\begingroup$ So in terms of innovation (excluding render times and performance) there isn't much left to do for offline rendering for light transport, does this mean offline rendering has hit a dead end? Would you happen to know of any of the challenges people who work on path tracers have? Light physics is what really interests me and the fact that the focus is moving towards animation really sucks :(... $\endgroup$ – Arjan Singh May 29 '17 at 17:06
  • 1
    $\begingroup$ Well actually I think there is more to figure out about basic lighting. Give this a read (Disney's PBR paper from 2012) to see some open problems with basic things: disney-animation.s3.amazonaws.com/library/… $\endgroup$ – Alan Wolfe May 30 '17 at 13:36
4
$\begingroup$

Since you've used the "path tracing" tag, I'm assuming you're thinking mainly about light transport. At SIGGRAPH 2015, there was a course on the topic of path tracing in movie production that may be interesting: https://sites.google.com/site/pathtracingrevolution/ Some of the presenters point out unsolved challenges towards the end of their presentation, for example see the second to last slides in the presentations from Alexander Keller and Johannes Hanika.

Then there are challenges in rendering that are not necessarily tied to path tracing but also present in game engines. Triangle meshes as an approximation of smooth solid surfaces are one, the use of shading normals or even worse, bumped normals have been pointed out as a source of artifacts a long time ago: "It's not really a rendering bug, you see..."

|improve this answer
$\endgroup$
  • $\begingroup$ I did check out the presentation, it said 'omission of effects results in images a little bit too dark' I'm a little curious as to what that really means. $\endgroup$ – Arjan Singh Jan 15 '17 at 9:17
  • $\begingroup$ It means several things: One is that path tracing cannot sample all possible light/scene interactions (see section 8.3 of Eric Veach's thesis). The other is that for production purposes, scenes are often rendered without caustic paths and a limited number of bounces (omitting those paths can significantly reduce noise). Terminating those paths removes light from the final result, making it darker than it would be if all light transport paths were calculated. $\endgroup$ – Stefan Werner Jan 15 '17 at 11:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.