NT Goes Head-to-Head with High-End Graphics Workstations
As Windows NT Server charges into the server and enterprise computing
market, Windows NT Workstation is quietly becoming a major force in computer
graphics. The release of Windows 95 has given rise to a crop of inexpensive and
accessible 32-bit 3D graphics programs. Until now, these same features required
high-end workstations and expensive, proprietary software.
3D modeling, rendering, and animation place tough demands on hardware and
operating systems, and the memory requirements and complexity of this class of
software can easily cause system crashes. Windows NT has become the ideal
platform for many graphics professionals because of its stability, memory
management, and multithreading capabilities, as well as its low cost and
versatility. You can render images and animation in the background of your NT
Workstation while you work with other graphics in the foreground.
Before NT, all but the most high-end workstations could be tied up for
hours while you tried to render a single image. The straightforward networking
capabilities of NT help with distributed rendering, meaning you can farm out an
animation to a number of workstations and each one can produce different frames.
These new programs bring workstation capabilities to the average user and
power and flexibility to experienced designers. In some cases, you can even
perform modeling, rendering, and animation in a single view.
Any one of these programs is suitable for product design, architectural
visualization, illustration, and short animation, and street prices are between
$200 and $500. I tested these programs on a 133-MHz Pentium system with 32MB of
RAM and a 2MB video card, minimum requirements for this type of work. Although
all the programs will run on a 486, they showed some strain even on the Pentium
with 3MB or 4MB projects. With the exception of Visual Reality, these programs
only run on Intel platforms.
TriSpectives
3D/Eye's TriSpectives is a solid modeler. It is based on the ACIS engine, a
new 3D file interchange standard developed by Spatial Technology, which makes
TriSpectives an attractive addition to computer-aided design/manufacturing
(CAD/CAM), scientific visualization, and games development. This standard allows
you to retain your models' spatial and physical property information while you
exchange them between programs.
The tutorials offer a good introduction to the modeling and placement tools
you use to build accurate models. TriSpectives has two metaphors for the
workspace: Scene and Page (see screen 1). Scene is a 3D workspace where you
build and render models and create animation. Page is a 2D view into a 3D world,
where you arrange models to create illustrations.
You can store your objects, textures, and animation movements in "catalogs"
and drag and drop them into a Scene. Users can even build their own catalogs so
that they can use elements in other projects. TriSpectives uses IntelliShapes,
which are basic geometric objects that have innate "knowledge" of
gravity and position. This knowledge makes it easy to use them as building
blocks for complex models. You can set anchors on objects to determine where
they align with or link to other objects. You can then modify or combine the
shapes with Boolean operations and Non-Uniform Rational B-Splines (NURBS), which
are curves with greater control and complexity than Bezier curves. You can also
modify the shapes by changing values in the property sheet, by dragging parts of
the model, or by using 2D drawing tools. You can drag and drop textures on
objects or on individual surfaces, and you can drop animation on individual
objects or on the entire Scene.
Two nice features of TriSpectives are its abilities to bevel an object's
edges from the property sheet and to hollow out, or "shell," an object
to give it an inner and outer wall. You can add Smart Dimensions for precise
placement of objects, but you are limited to technical drawings with these
dimensions because font, letter size, and arrow types are not configurable
except by editing the Registry.
You do all your work in a 3D window, which you can split if you want to
look at more than one window at a time. Actually, split windows are essential
because you can't save camera angles or views. Once you zoom, pan, or roll the
view, you need to manually enter its coordinates or play with the view tools if
you want to return to it.
You can create multiple scenes, but you have to individually populate each
one with the objects. You can also use TriSpectives in a 3D Page mode to create
a page layout with 2D and 3D elements. This is useful if you want to produce
illustrations and brochures.
TriSpectives will import and export Virtual Reality Modeling Language
(VRML) files, the new virtual-reality 3D format for the World Wide Web. You can
even configure the program with Netscape to act as a VRML viewer. TriSpectives
is Microsoft Office-compatible: This means you can embed data from any Office
application in the program, and you can embed TriSpectives data in any Office
application.
TriSpectives can quickly generate final output renderings. However, setting
up a final rendering is confusing because the dialog for defining output size
and resolution doesn't appear until after you choose File/Export Image and name
the output file. Other programs let you set the parameters first.
I found the program's initial performance painfully slow. I called the
company's technical support and they told me to change the Registry setting for
HKEY_CURRENT_USER/Software/3DEYE/TriSpectives 1.0/Raster/DeviceRaster Type from
5 to 6. Although technical-support representatives are helpful and
knowledgeable, you are billed on a per-incident basis after the first free call.
You can find free support and a 30-day trial CD-ROM on the 3D/Eye Web site. The
trial product is free, but you will need to pay shipping and handling.
Visual Reality
Remember those times you've opened a new software package and discovered
that the program was so huge it required seven floppy disks? Well, Visual
Reality 2.0 ships with seven CDs. The first one contains the program,
and the other six contain 2.5GB of models, textures, animation, video clips, and
complete projects.
Visual Reality includes a number of separate applications. The core is
Renderize (see screen 2), which was the strongest rendering product I reviewed.
Other modules are Visual Model, Visual Image, and Visual Font. Visual Reality
has a long pedigree of NT development, and you can see the results in the
program's speed and output quality. Even though it's not a ray tracer, Visual
Reality produces more realistic renderings than do the other packages.
The Visual Reality interface is unusual, to say the least. At first it seems
confusing, but excellent tutorials walk you through every nook and cranny of the
program. You'll be able to produce beautiful images after you complete them. The
keys to using the interface are three "wells," called View, Move, and
Edit. You can drop objects, views, lights, materials, and animation into the
wells, which will then bring up the modification options. With a little
practice, the system becomes very efficient. Users can control every aspect of a
rendering, ranging from how an image is wrapped around an object (mapping), to
the transparency and reflectivity of a surface, to the degree of fog in the air
and the density of the shadows. You can export your projects to VRML if all the
textures are .GIF or .JPG images.
The program offers alpha-channel masks for converting your 2D images into 3D
objects. For example, if you have a photo of a tree against a solid blue
background, you can make the blue transparent by setting blue as the alpha
channel. The leaves and branches will cast accurate shadows, and you have given
the tree the illusion of being 3D by setting the bump depth. You can create
elaborate, detailed renderings from simple models with this technique and can
populate them with realistic people, animals, and plants.
You've probably heard the adage: "A chain is only as strong as its
weakest link." Well, Visual Model is the weak link in the package. Although
it's a competent and precise module, all the basic tools and some unusual
features (including the surface-deformation tools in the Munge menu) make it
confusing. Even experienced designers will need to keep the manual close at
hand.
All the packages I reviewed allow you to extrude text into 3D forms, but
Visual Font is by far the most sophisticated (see screen 3). The well-designed
interface lets you instantly preview extrusions, bevels, offsets, and other
deformations. You can fit the text to a path and export it directly into
Renderize. You can return to Visual Font and edit the text as long as your
Renderize session is open. Visual Font comes with eight predefined bevels, which
you can invert and stretch. The information on bevels is stored in an ASCII text
file so it's available for user customization.
The animation features in Visual Reality 2.0 are vastly improved from the
previous version. Users can move any object, camera, light, or texture
individually or as linked items. You can assign separate texture and bump maps
to an object and animate them separately. For example, you can use one image as
a texture map on water for the color and pattern while you use another as the
bump map to create the waves. For some spectacular animation, users can set the
texture map to undulate slightly while the bump map moves diagonally. The manual
is thorough, and Visual Software offers free telephone technical support if you
do run into problems.