TITLE: Stolen!
NAME: Andreas Grates
COUNTRY: Germany
EMAIL: Andreas.Grates@public.uni-hamburg.de
WEBPAGE: -
TOPIC: Museum
COPYRIGHT: I SUBMIT TO THE STANDARD RAYTRACING COMPETITION COPYRIGHT.
JPGFILE: stole_ag.jpg
ZIPFILE: stole_ag.zip
RENDERER USED: 
    POV-Ray for Windows 3.61

TOOLS USED: 
    PaintShop Pro 7 (branding with sig & conversion to jpg)

RENDER TIME: 
    7 hours 31 minutes 17 seconds

HARDWARE USED: 
    AMD Athlon 650 / 320 MB RAM

IMAGE DESCRIPTION: 


"Red Herring stolen! Masterpiece disappeared from the New Zork Museum of Mobbish
Art!"
 -- New Zork Times, January 1st 2005
        
That's what the news was in the morning, this is what the site of the crime
looked like 5 minutes after the theft.
        

DESCRIPTION OF HOW THIS IMAGE WAS CREATED: 

        
Idea:
ad no real clue until beginning December. Thought about several collections of
things up until then, but found my modelling skills (& tools) getting in the
way. Since "museum" attracts this kind of pictures, and I wouldn't be able to
stand up to them in technical terms, I resorted to something you'd /also/ find
in a museum.

Lighting:
How to create the impression of night while still being able to see something
and not being thrashed and given a comment on the "poor lighting" again?
I used very pronounced radiosity and a dark blue shadowless light for the
ambience, and a light blueish area_light to emulate moonlight falling through
the dome in the ceiling directly on the site of crime. Photons were really
last-minute-extravaganza, since the render looked like taking ages anyway.
Overall, the lighting is as natural as nylon, but it makes quite a "nighty" feel
while you're /still/ able to see!
By the way: The laser beams are /not/ made with light & media. Would have take
too long to render, and I had the deadline coming.

Modeling:
Still using POV-Ray only, so there's really "only" CSG. The cactus is a recycle
from my first submission. I think I'll keep it as mascot!
The floor might be worth mentioning, since it is really made up of small boxes
individually textured by a macro.

Texturing:
Everything procedural!
A good deal of the textures is build from scratch (the pedestals for example, or
the glass), some are based on inc-files from POV-Ray. The least thing done was
putting together textures from inc-files in a texture map and applying a
pattern-function.

Special notes on this picture:
The laserbeams are delicate things. When applying antialias to the final render
I ran right into the open knife. Standard values made the beams look jagged,
ragged and ugly. I had to resort to pumping up the depth of AA to 7 and the
threshold down to 0.1 using the recursive AA method to get decent results. Ages
of rendering!
Postscript: Switched to non-recursive AA method. Performance dropped as low as 6
PPS in areas where nothing critical - well, just wall really - was! Really
faster now, but some ugly jags. Ugh. One of these days I should buy a new
motherboard, CPU and RAM.
Post-Postscript: Experimenting on another computer yielded that I could savely
lower the depth to 3 using non-recursive method. Even more speed, seemingly no
quality lost over the first third of the picture - which mainly contains the
laser-beams, the reason for this bloody experimenting! Hve to keep up the low
threshold, though, meaning count of supersamples stays high.

General notes on progress as artist:
Tried out some free modelling tools. Won't mention which since none really
suited me and I don't want to be rude about anyones modeller of choice. "Art of
Illusion" is currently in testing. Isn't really a modeller anymore, but has the
export and seems to meet with my intuition when it comes to usability. Java
though. Shouldn't use it for rendering.
Toyed around with arbaro (tree-generator) and planet Genesis (landscape
generator). Nice! Maybe I'll find a use for them!
Changed naming-scheme for submissions to irtc from "2-letter initials"_"5-letter
abbreviation" to the reverse to place the picture in a different position in
the listing. No real evidence on this, but I've got the feeling pictures
alphabetically at the front are judged harder because in standard ordering they
come first. If I'm bored enough some time I'll put the results of the IRTC re
relative alphabetical position/achieved score through the wringer. It should be
something like even distribution, but I suspect a slight tilt down in score in
the frontpositions.
No one commented on my new signature in the last submitted picture, even though
I asked for it explicitely. Possible inferences: Negative: No one reads this.
Positive: No one about to comment reads this.

