In article <dbngn5$265$1@chho.imagico.de>, chr### [at] gmxde says...
> Patrick Elliott wrote:
> > 
> > And once again, the existing method, while more accurate, is impractical, 
> > complicated and incomprehensible to most people for any reasonable 
> > application they might want to try it with. Even the people that have 
> > experimented with it so far have yet to get it to work as easily as it 
> > would need to be for a newbie to employ it. As someone that doesn't spend 
> > all my free time trying things in POV-Ray, that includes me. lol And, as 
> > you point out, its quite likely slower, which also makes it impractical 
> > for some uses.
> > 
> 
> What exactly makes you think that any such patch is easier to use?  On 
> the contrary - a patch that takes a less universal approach than a full 
> scattering media simulation to gain faster results most likely requires 
> additional controls to be set.
> 
Ok. The major problem I see is the memory, parse time and complexity 
needed. Now, with media, its probably impossible to 100% fix the render 
time. However, what about something simpler? Like:

subsurface {[image "blah"]
           [image map]
           depth <blah>
}

In other words, instead of having to stuff a skull or skeleton into the 
model, simple have a second 'skin' generated inside the object, which 
automatically creates an internal structure, where ever it can fit one? 
See, that is the advantage fake SSS has, the ability to arbitrarily 
define the 'depth' of the media and what ever physical characteristics 
the bones, etc. add, without having to make an entire model 'just for 
those things', and in the process maybe tripling the parsing time, 
quadrupling the memory used, etc. That's why 'bad' solutions, even if not 
realistic, are more useful. You don't have to spend 2-3 times as much 
time building a 'subsurface' for the light to actually scatter off of as 
you do designing the original model, which is the current problem. 
Without it, the media method doesn't work well and even with it, it still 
can't entirely approximate what is needed.

-- 
void main () {

    call functional_code()
  else
    call crash_windows();
}

Post a reply to this message