#REAL GLASS ?
16 messages in this thread
HELP !! All I want to do is create real glass in 3ds, you know the
stuff. Like when its a sphere it distorts the background and when its
lens shaped it magnifies. I`m pulling my hair out (hurry I don`t have
much left) Can YOU help ?
Since 3D Studio doesn't have a raytraceing renderer, you'll have to
resort to deciet and trickery to get the "warped" effect of looking
through glass. A combination of the right glass material (ambient 0,
diffuse 0-20, specular maxxed out, good auto or cubic reflection
map), and an animated texture map to simulate the warping behind the
glass. This is no easy task, but if you want to do it in 3D Studio
it's your only option.
To create the warped texture map you can create a camera with a very
wide "lens" that will warp the view for you, and place it in the
appropriate position in your scene.
Then apply this texture map to the faces, or create a fake background
that will serve as a "screen" for the distortion map, that will best
"play back" the glass distortion effect. It's not easy, and you'll
have to really plan your shots and motion to make this work, but it's
the only way I've ever done it.
Anyone jump in and suggest a better way, because this was a real pain
and I'd love to have a quicker, easier way to do it, too.
IMHO it sounds like you're trying to use a monkey wrench as a hammer.
If you expect realistic raytracing, use a program that is built for
it… you may end up with a few less dollars in pocket short term,
but a lot less grey hairs in the long run compared to the stress of
mediocre results all the way through (and the added grief of clients
who look elsewhere for the quality they miss).
GIG 3D-GO (running only on Silicon Graphics and/or SUN) at a cost
below US$ 10k offers raytracing of glass superior to Wavefront,
Alias, SoftImage, TDI, etc <all costing more in the range of US$ 35 –
50k>.
A system based on the new SGI Indy <worth looking at anyway> with GIG
will cost in the range of US$ 19k. and will raytrace output (stills
up to 8k or full animation) that will embarrass the owners of US$
60k+ systems. The GIG interface is logical, and the Indy will toast
any 3 networked machines for rendering.
We use 3DStudio <waiting for r3 to get to Hong Kong> for the quick and
dirty stuff, but when the client really wants to be impressed, we go
to GIG. No contest. Constructive Solid Geometry (CSG) means a ball is
a ball __no polygons to cause artifacting or mismapping__ and will
bend light (as lens) flawlessly.
If you remain desperate and can't justify the cost (even though the
Indy will run Windows/Mac emulation, comes with a digital colour
video camera on top, and performs internally *7 times faster than
Pentium*) I may be able to get you in touch with someone who can do a
GIG version of your output for you <either us here in Hong Kong,
GIG's hq in the Netherlands, or one of the users in the US in your
area>.
GIG will import .dxf, iges, texture maps and procedural routines for
"metaball-like" sculpting (fluid junctions of surfaces, including
custom texture/transparency blends while animating) and provides
links to an ever-broadening range of tools from physics rules for
NURBS
Best of luck with your wrench… if you'd like to try a hammer, let me
know. (or stock up on grecian formula and band-aids).
Digitally
Charles Watson
CLIC CADD Ltd.
69 Blue Pool Rd. Basement
Happy Valley
Hong Kong
Tel: [852] 891 4289 Fax: [852] 834 5658
Email: cliccadd@attmail.com CIS: 100314,655
We've seen GIG 3D-GO and it's output isn't anything to get carried
away with, as to your comparison with Wavefront, SoftImage and Alias,
your way off the mark, everyone of those programs specifically offers
tools that aren't available with the GIG interface and you would do
well to investigate such facts before you start attacking other
programs.
Additionally most of the core packages of the programs listed,
including SoftImage can be had for way under $35,000, so your
comparison is again flawed. I'd suggest you take the time to
investigate and get off of the stale rhetoric the dealer gave you.
Gig's lack of performance is the direct reason that most core
aniamtion houses and FX teams don't use it, but instead opt for other
programs.
As a person who has run all of the systems you mentioned on systems
from the Indigo to the Reality Engines, I can honestly say that Gig
was our least favorite and only shines when compared to
non-workstation units. Ala – your apples to oranges, again.
–> I'd suggest you take the time to investigate and get off of the stale
rhetoric the dealer gave you. <–
I apologise if my comments seemed to be "stale rhetoric"… we're beta testers
for GIG, and perhaps our inside experience of the features and bugs has
coloured our perspective on the issue.
–> your comparison with Wavefront, SoftImage and Alias, your way off
the mark, everyone of those programs specifically offers tools that
aren't available with the GIG interface <–
I've also had the opportunity to work with Wavefront and Alias (not
the newest releases, but recently) and readily admit that each has
its strengths and weaknesses. Alias has superior modelling [no
doubt], Wavefront offers better tools in Advanced Visualizer [at
costs to match] and SoftImage was ahead of many others in Metaball
technology and in their "flocking" heirarchical procedures for
grouped animation. Each could be considered a leader in their
respective bailiwick, but they sometimes suffer from their cumbersome
interfaces (GIG may be redundant in mouse-click %, but it is
consistent [training time shorter]) and $$$.
GIG is not intended to do all things better <and clearly doesn't>. I
have my pet peeves, and the clients we have resold the software to
have also passed on their beefs… however, we have been successful
in getting features changed and added {the original version of GIG
didn't offer .dxf or IGES import- now it does}.
But unless we wanted to increase our cost by 600 % (for example Alias
Power Animator's cost versus GIG), we can get a functional solution
that gives us a significant competitive advantage (not to mention
qualitative) over non-workstation software for a reasonable
investment. Days instead of weeks, real glass instead of blurred
reflection maps, and better price/performance are the motivations for
us to recommend it to our clients.
If I could find a better solution (in a comparable budget) I would
jump in and check it out with vigour. I keep my hand in wherever I
get an opportunity (chats with the developers and CEOs are often
better at revealing truth than conversations with the marketing
people… surprise <g>!) At this point, I'm also pushing Alias for a
client who doesn't know which package he needs, but has more
requirements in Alias' field than GIGs.
At the end of the day, expectations -both quality and cost- need to
be realistic and synchronous. For those who are limited by cash to
the PC world, 3DStudio shines (but doesn't raytrace). At the Reality
Engine level, assuming that cost isn't an object, one could chose
Wavefront, Alias, etc. and get stunning results which justify the
investment made due to their 'graphique verite'. Picking a winner in
the middle ground is a little bit tougher… GIG is better than some
and worse than others (and we're working on some of each).
I look forward to improvements in all of the software mentioned,
because I believe that these tools will form the foundations for the
imaging and visual media of the next generation.
<Remember, they put men on the moon with 8k… now people put that on
their wrist {for less money}!>
–> As a person who has run all of the systems you mentioned on
systems from the Indigo to the Reality Engines, I can honestly say
that Gig was our least favorite <–
OK, the obvious question would be… what was your favourite
(hardware/software/solution) and why?
It's valuable to find someone who's been there (since there isn't much
support on CI$ for SGI tools).
I'd be curious to correspond with you further on what you see as the
(dis)/advantages of each through your exposure to the high-end world.
Are you strictly a user of these programs, or do you consult and
pitch for your favourites? Do you agree or disagree with my
suggestion that people look into INDY? Can you fill me in on ANIM8
INC?
I try extremely hard not to discuss alternate programs on the Asoft
forum, but there will be an e-mail to you with our
perspectives/outlooks on the SGI software market, where we think it's
headed, personal favorites, and who we are.
I try not to step on any toes (don't sell Chryslers in the Ford
dealers face), but there are times where I shake my head a bit and
wonder why a little broader exposure isn't more common… after all,
the world is coming online to exchange information (amiga forum folks
comparing Lightwave et.al.), and with respect for individual opinion
and asbestos-lined attitudes, there's no reason not to provide people
with open access to information.
After all, isn't confidence about your product supposed to allow you
to accept minority opinion?
FWIW, most of the clients and co-workers I've contacted are diehard
3DS fans (without seeing R3 yet!) and are secure in their faith that
it will continue to dominate the PC world… Great Work and great
expectations!!
Where did you come up with this idea that the Indy performs "*7 times"
faster than the Pentium? The better Indy model comes with the Mips
R4000SC, which is rated at about SpecInt 60/SpecFp 64, while the
Pentium is rated at SpecInt 62/SpedFp 59. I've seen other floating
point benchmarks that show the R4000 can be somewhat faster on FP
than the Specmarks would indicated, but not 7 times faster, not even
2 times faster. The base Indy comes with the R4000SC, which is
somewhere between 486 and Pentium in Performance. The R4400, which
is available only in the more expensive members of the SGI line is
considerably faster, but still no 7x.
Maybe you talking about the relative speeds of the memory busses? I
would believe that the Indy's memory system is n times faster than
the typical pentiums machine's, but I don't know if I'd accept 7 as n
without proof. The Indy is optimized for multimedia type digital
video, but memory bandwith is not nearly as important as floating
point performance for rendering, especially the raytracing sort of
rendering done by 3D-GO.
I've seen the still output of 3D-GO, and it is very impressive. I've
played around with converting 3DS scenes to POV-Ray format and
raytracing them. I really like the way they look. It would be nice
to have a fast raytracer like 3D-GO.
I was quoting the 7 times number for the internal DMA transfers of the
Indy.
Scott Bonham (one of the managers of the INDY development at SGI) had
listed this figure for the internal 400Mb/sec transfers and 267Mb/sec
throughput versus the capacity of the buses on the pentium boards
which are currently available and in comparison with the new Mac
Quadra 840AV.
His parallel was that of a fast engine placed into a cheap drivetrain
and feeble clutch… lots of noise and great performance for the
pistons, but when you want to put that power onto the road, you're
more likely to just drop the block (can you say "seizure") or blow
your gearbox than get true output of that horsepower.
I didn't mean to mislead anybody with performance figures, but I've
seen enough Ferraris that spend their whole life in first gear due to
traffic jams to know that you need more than a chip to get
performance.
With regard to GIG raytracing… I can't say much, but the pipeline
talks about _Parametric Raytracing_ as the next feature to wow…
interactively evaluating scene info so that the redraw only applies
to the areas of the action that are altered… <don't quote me yet,
but save a few pennies for the day it ships>
Give me a rundown on your setup and your wishlist… Xmas is coming
and maybe I can talk to Santa… alternatively, I may be able to
throw a suggestion or two your way.
Charles,
The "Ferrari engine with Pinto drivetrain" analogy is valid for many types of
computing, but I don't think it matters much for floating point bound stuff
like 3D rendering.
An example: once upon a time there was a PC called the ALR Powerflex 286. It
was sold as a 12MHz 286, but could be upgraded via a CPU module to either a
386sx/16 or 486DX/25. With the 486 module installed, it was essentially a
25MHz 486DX with no external cpu cache and with the memory bandwith of a 12Mhz
286 (80ns 16bit DRAMS). Byte found that it had integer and memory benchmark
performance barely on par with a 386/25, but on floating point stuff it was
about 90% as fast as a Compaq Deskpro 486DX/25 system.
Floating point is limited by the internal speed of the CPU, not by the memory
bus bandwidth. There are darned few useful things you can do with data at
400MB/sec besides copy it. But this is what the Indy was designed for:
sloshing video data around in main memory. You need lots of DRAM memory to make
this useful, otherwise you're back to paging from disk and memory bandwidth is
meaningless.
Personally, I'll wager that the 66MHz P24T (Pentium upgrade chip) in a 486
motherboard will render almost as fast as a full tilt 60MHz Pentium.
I'm definitely a small time operator. I have one machine, a 486DX2/66 with
32MB ram and an 877MB hd, Targa64+, 4mm DAT, CD. There's no way I could afford
an SGI system, and my next major purchase will be a cheap Pentium box.
Bailey
Seems to me there's a file in this forum called AMR.TXT [Advanced Modeling and
Rendering] that describes a book/software deal. I think the software called
Mirage, will raytrace a .3DS file. There was also something about if you buy
the book now, you'll get a break on the final version of the ray tracer.
This might be exactly what you're looking for. Good Luck, Eli.
There's actually an easier way of doing this than JohnT mentioned.
(there's an example of this technique in the last shot of The Voyage
on the 3DSr3 Siggraph demo reel)
First, render a matte of the glass. Then, distort the background with
an image processing routine (blow it up a bit, blur it out, etc). Use
the matte to reveal just the distorted background behind the actual
glass and the effect will be perfect.
Thanks, Gary. When I did the backflips for glass distortion, mattes
weren't available (R2), but I suppose I could've figured some VP
alpha trick out instead. This sounds like a much better way and will
be filed away for future reference.
I was sitting reading your message about 20 times to try to
understand the technique you were describing about using a matte, etc
to get a great looking refraction-glass effect.. but I couldn't quite
figure it out. I didn't want to write this message in fear of
missing something too simple or being sent to my reference manual
(which I have done looking up for "matte").. but I can't put 2 and 2
together to achieve this refraction effect. Would it be asking too
much to get you to more or less spell out some of the important steps
to accomplish getting a refraction effect (say of a panning camera on
a wine glass) again? I'd really _hate_ to miss out on this handy tip
for future reference.
I have never done this, but this is how I understood the process(in
simple terms)-(Please, anyone, correct any mistakes I make here!)
1. Complete your entire scene with all camera movements.
2. Hide the 'glass' object, so that you have an unobstructed view of
whatever is behind the 'glass'.
3. Render the animation, effectively giving you a moving background.
4. Take this animation to an image processing program(ie.
Photostyler?) and alter the 'look' of each frame, so that they are
distorted (magnified, blurred, etc.)
5. Go back to your original animation, unhide the 'glass', and use
the deformed background with a matte (sized and shaped to fit the
area affected by the 'glass') so that only that portion of the
deformed background shows up when rendered again.
Like I said, this is just my understanding. Is this any clearer? Good
luck!
BobW's outlined it perfectly.