#Refraction?
19 messages in this thread
There is a competition on my local BBS between the guys using POV
against the guys using 3D Studio. The competition requires that we
create the most realistic image of a bottle of wine. POV has the
slight advantage in that it can do the refraction. How could I
simulate refraction in 3D Studio? Let me know if you have any tricks
I could use. We couldn't let POV win could we!
You should use some of the available 3DS add-ons (for e.g. RayMan)
routines to get a realistic image. If you can`t get a hand on RayMan
fast enough you might find a RayMan beta user to render your scene…
I've never done it, but I hear you can simulate refraction by
applying a texture map of the background to the wine bottle. Place a
camera with a wide angle lens at the spot of the bottle and capture
an image of what's behind the bottle. Then apply it to the bottle,
or perhaps a transparent copy of the bottle. You'll have to flip the
image along the x and/or y axes to get a realistic effect, I believe.
The "3D Special Effects" book has examples of how to achieve refraction.
– J
PMJI
Gosh, I really like the special effects book and have learned alot –
but – the illustration of the wine glass is not going to win any
contests in competition with a ray tracer. I would guess that the wine
bottle subject was suggested by on of the POV guys since refractions
are the only area where POV has an advantage. The really impressively
creative complex methods 3DS users have come up with to simulate
refractions is an indication of the need for selective refraction
mapping. Ill probably get Rayman for still work but even if it
supported animation it would probably be too slow for that kind of
work. Maybe an animation capable Rayman , travelling alpha, and
videopost would be the closest workaround.
I agree fully. I am a beta-tester for Rayman and I really don't see
point in porting my scenes for any other reason except refraction. It
is so slow, even on my P-90. The main problem is that you cannot
transfer your IPAS stuff to Rayman, which is what makes 3DS so good.
Even if RM had animation capabilities, I would't pay the price for
it. You will be able to by a complete package (Lightwave for
Windows) soon for only a little more money.
Back to my point. What 3DS really needs is some kind of refraction
mapping , either built in or via IPAS. It is the only real weakness
in the renderer. Some 3rd party developer could really make a
killing on such code.
Anyway, just a thought.
Well, did you see BATH.JPG in the images lib? Forget the glass and try
to make a image with the same quality in 3DS. You won`t get the level
of reflections and of course you will not get this kind of shadows.
3DS is not able to render penumbra and umbra for shadows. And the
anti aliasing. Did you see it?
If RayMan is too slow you probably used some unneeded options. You won`t find
a faster raytracer on PC`s.
>>You will be able to by a complete package (Lightwave for
>>Windows) soon for only a little more money.
Yes, and you can get Truespace for a little bit more. But what do you
want to do with it? Rendering? <bg>. I took a look at Lightwave but I
can`t say any details because of the NDA (you have a RayMan NDA,
too). But regarding the render speed and the image quality: <g>.
I agree with you regrading the IPAS. You can work around by using
alpha rendering and composing the image (Bill Batchelor did a great
image where he used vapor in a RayMan rendered image). But you`re
right: we need the IPAS. But let us wait for 3DSR5! A lot of things
will change…
BTW: take a look at your NDA.
Hello Michael,
On 28-Nov-94 you said to Mitch Gates:
->You won`t find a faster raytracer on PC`s. (Speaking of Rayman)
With all due respect…
Can you really, truly, prove this? Have you actually benchmarked a Rayman scene
against the registered (watcom-compiled) version of Vivid, POV, or Polyray?
Be careful what you say, this could easily be tested using the 3DSPOV18
converter.
-Heinz.
Hello Heinz,
>>Have you actually benchmarked a Rayman scene against the >>registered
(watcom-compiled) version of Vivid, POV, or Polyray?
Yes, we really have. And we tested it against Real 3D and other packages. Get
a RayMan copy, test it on yourself and if you find a raytracer wich renders a
scene of, let`s say 100,000 polygones, faster then RayMan, I`ll never say it`s
the fastest raytracer on PC.
But please run both on the same machine and the same configuration and render
options. I think a Pentium with 32 megs will do the job good. (I made the tests
using a Pentium/66 with 64 MB RAM and scenes from 20,000 to 250,000 polygones).
Don`t get me wrong, I don`t want to blame any software. I used POV for a long
time, it was my first raytracer, and I loved it very much. It`s really good and
gaves us lots of ideas to implement in RayMan. But when someone says that
RayMan isn`t fast than I can only say that raytracing needs of course more time
than Z-Buffering. But if you compare between different raytracer you can`t say
that RayMan is slow.
Hey, and by the way: I think it`s more important to get all 3DS features
(mappings, light settings and so on) than the point that A needs 5 second less
than B to render a scene. Only when A is two times faster than B you should
spend your time on making some benchmarks.
Thanks.
Michael
>>if it supported animation it would probably be too slow for that >>kind of
work.
That`s the reason why the first release of RayMan can`t render
animations. The second one will have a combination of a Z-Buffer (or
A-Buffer) and raytracing/radiosity. With this features RayMan is as
fast a other Z-Buffer but has the image quality of raytracing.
BTW: if you have some experiences with the RayMan settings (like
internal reflections, pixel-sub-division, softshadow turbo or not)
you can decrease rendering times for more than a few hundred percent.
RayMan is faster than 3DS with automatic reflections and raytraced
shadows. And it needs much less memory.
Any ballpark ETA on the second version? Will it support IPAS effects?
I think R2 will come out somewhere in summer 95. It will support 3DSR5 and I`m
not sure how the R5 IPAS will look like. But I4m sure we`ll support them.
Michael
Have you considered what the upgrade from R1 to R2 might cost ? The reason I
ask is Im trying to decide whether to get R1, which
would meet some of my needs or wait for R2 which sounds like a much more
complete solution. If the upgrade cost were to be a
very significant percentage of the cost of R2 then R1 wouldnt be as attractive.
Well, that`s a good question. I think the update won`t be very expensive
because we know that the most customer get RayMan to render animations. I think
a price around US$ 100-200 is very realistic.
I believe it`s a good step to get the first release of RayMan because it needs
a little learning time to render with raytracing instead of Z-Buffer. The
behavior of lights and materials are a little different.
Michael
Thanks
>> PMJI
Gosh, I really like the special effects book and have learned alot – but – the
illustration of the wine glass is not going to win any contests in competition
with a ray tracer. EDWARD HARVEY<<
Ed,
Glad you liked the book. As to the quality of the pseudo-refraction vs a
raytraced refraction: absolutely. Refraction is much too complex too allow a
manual simulation to match a raytraced simulation. The aim of the article was
to show that you can give the _impression_ of refraction using 3DS's own tools.
For many images, where the refraction is not the central area of interest, the
impression of refraction is sufficient and the savings in rendering time can be
important. Where the refraction _is_ the focus you really do need a raytracer
and will have to spend the time that a raytracer needs.
Terry Gilbert
Forcade & Associates
I appreciate the creativity of the very complex work arounds that you and
others have come up with in trying to simulate refraction.
My point is that these very efforts demonstrate the need for a refraction
solution for 3DS that dosnt negate the advantages of scan
line speed. To me this is the weakest aspect of 3DS which is otherwise by far
the best 3D software I will probably be able to afford.
Hi Terry,
I agree with you. Simulated refractions are very good for animations where the
refraction is not the most important thing in the scene. For stills where you
have enough time to examine every part of the image or animations wich lives
from the refraction you have to use a raytracer.
Michael
Here is a trick that was once posted by Grant Blaha. It might be useful for
you.
——–
I did an image that simulate refraction once, but you may not want to
replicate the method. I only did it with a single still image.
I rendered the scene's background geometry without the glass sphere. I
then created a flat mesh using grid_i.pxp and texture mapped the background
rendering onto the mesh. I then pulled the vertices of the mesh outward in
a spherical shape to create a distortion. By using texture bluring, I was
even able to give a glassy blur to the distortion.
I rendered an OPAQUE sphere without any other scene geometry, and kept
its alpha channel (using the seperate alpha channel option).
I rendered the entire scene. I inverted/negated the opaque sphere's alpha
channel and used it to cut a hole in the scene. I then composited the
distorted rendering into the hole.
Clearly, ray tracing is a simpler way to do it. An IXP could do it real
good, but I don't know if there are any. Any developers reading this should
be aware that R4 will provide access to the Z buffer, which would make this
effect EXTREMELY easy to do well.
Ciao,
Grant
——–