Hi everyone

Anton and I have been using radiosity in 3.1 for quite some time, see
www.antonraves.com for some examples and though they are not the best
setting possible, they look quite good and they are really fast. I
played with the included 3.5 settings yesterday a bit and difference is
rather big. First it takes a lot of changing in all the textures from
ambient to diffuse, which is something I can live with. But the render
times are very very slow. A picture that would take 30 minutes in 3.1
takes more than 7 hours with 3.5, and the result isn't even better.

Is there somewhere some information on how to change settings (even the
3.1 settings in 3.5 are dead slow) and how to get some decent speed.

TIA


Thomas

Post a reply to this message

Thomas wrote:

> Is there somewhere some information on how to change settings (even the
> 3.1 settings in 3.5 are dead slow) and how to get some decent speed.

I think that it's better to forget about 3.1 radiosity settings and start
anew. There are various examples in the current distribution (see the
rad_def2.inc), including in the advanced scenes (see for instance the
stackerday.pov and stackernight.pov scenes). Right now, it seems that
people who have experimented with it have their own preferred methods. Kari
Kivisalo, for instance, has done a lot of work about matching the settings
with real-life lighting parameters (see his posts in povray.binaries) with
good results. Other people have an approach based on trial-and-error. For
me, presently, the setting up of radiosity involves three basic (almost
independent) steps :

- a quality/speed/memory optimisation where I try to find the optimal count
and error_bound combination, both for testing and for final renders. I
usually end up with high count (>300) and low error_bound (<0.05) values,
but it heavily depends on the scene. Particularly, scenes involving large
expanses of flat, non-varying colours mixed with more complex material are
still a problem. AFAIK, small error_bound are the only way to get proper
shadows in the corners.

- a colour intensity optimisation where the brilliance/assumed_gamma/light
intensity (if any)/environment colour intensity come into play.

- a "colour bleed" optimisation where the gray_threshold/light colour (if
any)/environment colour(s) are tuned

I barely touch the other parameters, using pretrace values = 1 and
recursion = 1. Higher recursion values give more realistic results but I
find the render times and the artefacts unbearable.

And yes, render times can be incredibly long and the memory use can be
extremely high. Some of the new images done for my website took several
weeks to render and ate more than 700 Mb of RAM.

G.


--

**********************
http://www.oyonale.com
**********************
Graphic experiments
Pov-ray gallery

Post a reply to this message

That's strange. I have experienced the exact opposite.
  For example this image: http://iki.fi/warp/pics/Radiosity_test2/rad1_2.jpg
took about 1 hour to render with pov3.1 at 320x240, radiosity recursion
level 2.
  The same image takes with pov3.5, with properly changed radiosity settings,
less than 10 minutes to render at 640x480, radiosity recursion level 4. The
result is practically the same (although the higher recursion level adds a
bit of illumination here and there).

-- 
#macro N(D,I)#if(I<6)cylinder{M()#local D[I]=div(D[I],104);M().5,2pigment{
rgb M()}}N(D,(D[I]>99?I:I+1))#end#end#macro M()<mod(D[I],13)-6,mod(div(D[I
],13),8)-3,10>#end blob{N(array[6]{11117333955,
7382340,3358,3900569407,970,4254934330},0)}//                     - Warp -

Post a reply to this message

> And yes, render times can be incredibly long and the memory use can be
> extremely high. Some of the new images done for my website took several
> weeks to render and ate more than 700 Mb of RAM.

Well I guess we need lots of time then and experimenting. Maybe it is a good
thing I managed to get 64 bit version compiled, then we can use more then 4
Gigabyte ;)

BTW I looked at your new image the other day and I'am most impressed! Excellent
work, I wish I could do stuff like that.


Cheers,

Thomas

Post a reply to this message

Thomas wrote:
> BTW I looked at your new image the other day and I'am most
> impressed! Excellent work, I wish I could do stuff like that.

I also take a look at them, and... well, words are just not enough.
Gilles, the girl in "Birds and bees" is not rendered, is she? If the
picture was made by anybody else, I'd say that couldn't be rendered, but
when talking about Gilles, you never know...

-- 
          /"\                           |    iki.
          \ /     ASCII Ribbon Campaign |    fi/
           X      Against HTML Mail     |    zds
          / \

Post a reply to this message

Jari Juslin wrote:

> I also take a look at them, and... well, words are just not enough.
> Gilles, the girl in "Birds and bees" is not rendered, is she? If the
> picture was made by anybody else, I'd say that couldn't be rendered, but
> when talking about Gilles, you never know...

It's Gilles. It's rendered. Poser figure, probably. Just with the magic
touch of Gilles on it, life jumps in.

--
Wind the Frog!

Post a reply to this message

Jari Juslin wrote:

> I also take a look at them, and... well, words are just not enough.
> Gilles, the girl in "Birds and bees" is not rendered, is she? If the
> picture was made by anybody else, I'd say that couldn't be rendered, but
> when talking about Gilles, you never know...

In this case, the texture is a very detailed image map (based on a real-life
human model *) on a very detailed Poser model (also derived from a human
model). I can't take any sort of credit for these as it's basically a
photograph pasted on a mesh. When I started working on this idea, in 1999,
neither the model nor the map were available, so I put the development on
hold until DAZ released them the following year.

The interesting thing is that in spite of the quality of the model and
texture, it's mostly the radiosity lighting that gives the realistic feel...
this and avoiding to show the character's face and the original flesh
colours, that just didn't look right.

G.

* no, I didn't took the pictures...
--

**********************
http://www.oyonale.com
**********************
Graphic experiments
Pov-ray gallery

Post a reply to this message