13 years ago, at SIGGRAPH 98, Softimage announced its next-generation rendering system, code-named “Twister”:

First Component of “Sumatra” Provides Users With a Glimpse of the Future of 3-D Rendering

ORLANDO, Fla., July 20, 1998 — Softimage Inc., a wholly owned subsidiary of Microsoft Corp., announced today at SIGGRAPH 98 that its next-generation rendering system, code-named ” Twister ™,” is entering the beta test cycle with delivery to customers scheduled for the fourth quarter of 1998. “Twister” is the first module of the soon-to-be-released 3-D animation platform, code-named “Sumatra,” which provides a revolutionary new approach to character animation and introduces the concept of nonlinear animation (NLA).

Twister was considered at one time to be the “the first step toward the integration of 3D in Digital Studio”, and was basically the rendering component of Sumatra (XSI).

The last-ever version of “Twister” was beta 2.5, shipped in June 1999. By then, it had been decided to combine “Twister” and “Sumatra”, because “Twister”, as a standalone product that relied strictly on .mi2, posed too many workflow and performance problems.

Here’s the executive summary from the Twister v1.0 Product Brief:

Twister v1.0 is a standalone rendering product based on the Sumatra Architecture, designed to work in conjunction with Softimage v3.7SP1. The goal of the product is to provide a finishing and rendering solution which features next-generation render functionality , while providing a transition platform for our existing customer base onto the new generation of Softimage products. Twister fits neatly into the Softimage 3D v3.7SP1 workflow, providing an interactive toolset which complements and extends existing functionality.

Twister rendering is based on mental ray v2.0, incorporating a host of next-generation rendering features such as distributed tesselation, global illumination (radiosity), and caustics (light diffusion through transparent substance). Though Twister is not generally designed for the animation scene objects, it does provide full animation for any rendering parameters, as well as lights and cameras. All standard rendering properties may be applied, including shaders, motion blur, antialiasing, etc.

Twister imports scene data from Softimage 3D in several different ways. Firstly, Twister can read and use standard Softimage v3.7SP1 scene files. Secondly, Twister can work directly with .mi2 format. The mi2 data format is the next generation of the current mental ray mi format , used by Softimage 3D v3.x to communicate with mental ray. The mi format is actually a ‘stream’ of completely resolved geometry. The mi2 format greatly improves compression by only including what has changed between frames, rather than what is included in each frame. Twister will include a plug-in for Softimage 3D v3.7SP1 which will allow direct export of .mi2 format to disk. Twister will also include converters from several other industry standard formats (IGES, DXF, etc ).

Twister also contains an interactive plug-in for SI3D v3.7SP1, which allows users to dynamically ‘export’ their current working scene to Twister, and continue working on their scene in Twister (note that any changes made in Twister can not be propagated back to Softimage).

The Twister toolset is divided into two suites – the Main Suite, which includes scene and render viewing, interactive tuning, and light/camera manipulation, and the Renderfarm, which is used for distributed render control.

Now for some screenshots…

Here’s a mock up of the Twister main suite (back then, the term “layout” wasn’t being used to describe instances of “DS Suites”). Twister was originally spec’d to include three suites: Main, Renderfarm, and 3D Paint.

Mock up of the Renderfarm suite. The non-linear editor (NLE) at the bottom was intended to show rendering passes, and the mock up doesn’t show that yet.

Snippet from the beta documentation:

Screenshots of the Render Tree and Schematic View from the beta Twister docs.

Thanks to Darrin Hofmeyr, for sending in these screenshots from the SOFTIMAGE|Behavior section of the Production Series DVD:

First released back in November 2002 , Softimage|Behavior was a behavioral animation system for creating characters and choreographing 3D animated crowds that can respond to objects, other characters or changes in their environment. From the press release:

Softimage|Behavior lets an animator control individual characters in a simulation through a combination of graphical editing, scripting and debugging tools in addition to a library of pre-defined behaviors. Individual characters can be set to respond and interact in multiple ways to their environment and with other characters. The system’s integrated character engine features real-time IK, animation blending and warping, dynamic path planning, and automatic obstacle avoidance, which supports event-driven responses for characters in the simulation.

Softimage|Behavior is priced at $14,995

SOFTIMAGE|BEHAVIOR was used on the history-based mini-series Napoleon for the A&E Channel, and in James Wong’s feature film The One.
It was also used by Buzzimage for a Dollar Bank commercial. Here’s a collage of a few AVIs about this commercial that I found lying around:

http://vimeo.com/26475107

And there was a press release on xsibase too:

SOFTIMAGE HELPS MAKE MILLIONS FOR DOLLAR BANK
–Buzz Image relies on SOFTIMAGE|BEHAVIOR to populate the streets of Pittsburgh in the newest Dollar Bank commercial —

And finally, courtesy of the WaybackMachine, here’s the case study that was posted on softimage.com back in 2003:

Softimage customer story from 2004/2005 for Drop, a television spot for Absolut Cut Vodka, done by The Mill. See the spot here.

Created for Absolut and McCann Erickson NY by The Mill New York, the striking spot relies heavily on the intricate CG work of Swedish artist Stefan Andersson, a long-time aficionado of SOFTIMAGE|XSI.

The “XSI flower” was a graphic design element used with releases 4.0 and 5.0.

I call it the “flower” because I found it in a folder named “XSI_FLOWER”. It appears to have been done in Photoshop (there was a .PSD file there) and this JPG:

UPDATES:
Thanks to Hirazi Blue (in the comments) and takita (XSI list) for pointing out that the flower logo was created by me company using Softimage|XSI itself.

In the comments, Hirazi also mentions that:

There was a nice 5 page article specifically about this design in the August 2007, #93 issue of 3D World (pages 44-48).

So, I dug up (aka googled) some more info:

• The cover page for the XSI 4.0 New Features document includes this:
  

  Cover: “Water Flowers” designed for Softimage Co. by Me Company, using SOFTIMAGE|XSI.

• On xsibase I found this quote from Gareth Morgan (Senior Product Manager, Softimage):
  

  “We’re really proud of the new symbol. Yes, of course – it was 100% designed and realized in XSI. The amazing designers at Me Company (London, UK)worked with our Creative Director and designed the ‘Water Flower’ exclusively for v.4.0 – it’s a strong, colourful symbol that is easily identifiable, translates from print to screen beautifully, and communicates the vision of our brand for XSI. Me Company designed the Water Flower as model in XSI, and used the lighting and texturing tools in to give the flower it’s look and feel, and then used the advanced rendering capabilities in XSI to output the design.”

• The “water flower” logo was mentioned in the book More than a Name: An Introduction to Branding:
  
• Some info about Me Company:
  

  Although the majority of Me Company’s output is in two dimensions, the production process involves three, “XSI is our illustration tool of choice,” says Paul. “It’s the way we express our ideas.” The results, which Paul calls ‘hi-res stills’, capture the world of possibilities Paul and his creative allies envision in response to a particular brief. What makes this work is a faith in the value of the creative process itself: “We follow a line that interests us and try hard to push things forward.”

http://vimeo.com/25238551
Softimage|XSI The Bottom Line

From “Fuel For The Mind”, edition 1, La Cite des Enfants Perdus—Buf Compagnie 1995
(aka The City of Lost Children)

http://vimeo.com/24917633

I remember watching this back in the old days when I worked at Softimage/Microsoft.
Many thanks to Miquel Campos for giving me this CD.

What’s this flea made of anyway? It’s not an electronic model, not a digitized photo, it’s a cute little chimera concocted in C. G. I., modelled, dreaded, and animated by his creators at the Buf Compagnie, using a mythological method of mathematics that took hundreds of hours on the computer.

The surface texture was obtain-ed by using several layers of lovely materials such as beach pebbles, green beans, snake skin, pig skin, leather, etc. The animators studied hours and hours of documentary films to find out how fleas act out there in nature. Now they know everything about how fleas hop, skip, jump, land, sting and drink, and do scores of dirty itchy tricks that have earned them a worldwide reputation.

Given the variety of shots where the flea would star, the animators modelled a series of fleas with varying degrees of definition. The most sharply defined had no less than 800.000 polygons.

Prix Ars Electronica 1996

City of Lost Children trailer

Eye-Candy Reel 2005
SOFTIMAGE|XSI 5.0

http://vimeo.com/24621051

While I was cleaning out my basement, I found Computer Graphics World, Feb 1998, sitting on top of my pile of decades-old Sports Illustrateds.

On the cover, Geri from Pixar was created using a new surface modeling technology called “subdivision surfaces”.

In the Spotlight, “Maya…appears to have positioned itself as a true competitor with the likes of major players such as Softimage…”

Click to view full page from CGW

Formerly code-named Digital Studio, Softimage/DS is “a complete turnkey system [that] costs approximately $100,000”

Click to view full page from CGW

Feature article: Unraveling Riven
To achieve the “next level of graphic reality…Cyan switched from Mac-based Vision3D from Strata, which was used to develop Myst, to the SGI version of Softimage 3D.”

This huge set, which consists of more than 2 million faces and 20,000-plus models, required two hours just to load in Softimage.