Subscribe for FREE

Discussion Area

VRML by the Pound

Tell a Friend
We'll Link to You!!

Aereal, Inc.
   Instant VRML
   Proteinman's Top 10

view me!

The Challenges of Creating Content in VRML -- A SIG Report
by Sue Wilcox

Tuesday, April 18, 1997. Palo Alto, CA. The VRML SIG was back again in SGI's Cafe Iris to enjoy artistic inspiration from ThinkFish and Construct. This session was a balance of art and technology with both companies acknowledging a debt of inspiration to Scott McCloud author of 'Understanding Comics'. ThinkFish was represented by Ben Calico, Director of Creative Development and Chris Lowery, VP of Business Development. ThinkFish have the mission "to popularize non-photorealistic 3D rendering" according to Ben. If you're wondering what this means, it's all to do with letting your imagination do a lot of the work rather than having the rendering engine laboring away. ThinkFish use 3D, stylized via their rendering technology, so objects appear as if drawn by an artist, not a computer.


As with cartoons and artistic impressions, ThinkFish want your imagination to fill in the details: "the power of the imagination is greater than anything you can create to fill in a scene". They believe they have a stronger power to reach the emotions using cartoon/comic characters than with real ones. They are maximizing what is effective in telling a story. And by using 'living cartoons' as characters in their worlds ThinkFish can create a 'cartoon comfortable environment' that will not date as technology improves.

The ThinkFish engine analyses a 3D object and finds the salient lines - the lines that really capture the essence of the object - just as a cartoonist would when sketching a caricature or an artist would when drawing a model. These lines are an interpretation of the underlying geometry of the object and permit effects like overdrawing of corners to give a sketched look, use of different thickness of lines to give a pencil or ink drawn look, and the adoption of any style to produce a drawing of an object in that style. The styles can be Picasso, Monet or caveman - you pick. "We are very forgiving of what would otherwise be very crude models" said Ben. Because the lines chosen are a simplification of the geometry, the file sizes are smaller than regular VRML files and low polygon objects appear looking much better than they would do as just colored geometry. ThinkFish tend to avoid the use of textures so this keeps their file sizes down further. By avoiding chasing the Holy Grail of photorealism, which eternally retreats to the mists of improving technology, ThinkFish allows designers to actually complete their vision. Art can be achieved and completed and won't go out of date the way once cutting-edge quality images do when rendering technology improves.

Ben explained that the business model ThinkFish are using is to license their technology to companies (they've already done deals with Apple and Fractal Design), then to sell individual styles to users who want to turn out 3D images with a particular look. ($30 per style direct from the ThinkFish Web site.) If you want characters that look crayoned you would buy their crayon pack. Chris brought up some demos of their intelligent rendering engine drawing Homer Simpson in animated 3D, and then a cookie monster. Ben drew our attention to the line detail around the monsters armpit: every time the arm moves to a certain position the armpit crease line is added. "This is not minor science," Ben said.

ThinkFish runs on any Pentium or PowerPC machine and its small file sizes make it great for use over the Internet. We had a look at a 150 polygon VRML fish cartoon which has a file size of 23K uncompressed. With another 5K of animation performance or interactivity this would make a great character Ben told us. It avoids the giant pixel problem - because you can zoom in without image degradation - it avoids the need to use giant texture maps to simulate realism (because that's not its goal), and it avoids the giant machine problem - because you don't need lots of CPU and memory to run these small files. The ability to zoom in and out was something Ben gave heavy emphasis to: loss of picture quality when you try and get close to an object ruins the immersive effect of VR experiences, the use of vector lines which are totally scaleable prevents this loss of the suspension of disbelief. The alternative is to use LOD texture maps or FlashPix tiles both of which will increase file size. Ben points out "We can't count on the pipes getting thick for a while yet. Lo-res images download fast and perform well on lots of different machines."

Questions came in a rush of enthusiasm: Can it handle color per vertex? Yes, the processor has an abstraction process which means you can do almost anything to the incoming information: you can walk around the color space, color per vertex, parameterise transformations over time. "Its like a scripting engine inside the renderer itself - it allows you to do anything you can develop a style for." How large is the application itself? About 300k at the moment although that's expected to shrink as they work on it. They expect 5 to 10 k per style. It does add size to the model, but as it gets rid of the need for texture maps they feel that's OK.

How large are worlds built using ThinkFish? They don't really do worlds much at the moment, just individual characters. All Ben could tell us was that the Monet waterlilies scene they showed us was around 1,500 polygons. The scenes can be accelerated by hardware because the renderer is a triangle processor. Effectively ThinkFish scenes run at about the same speed as everyone else's VRML scenes - their claim to fame is being first in the real-time, cross platform, non-photorealistic market.

Is there a cartoon model maker tool? Only an internal one, they have no immediate plans to produce a character animation tool. They prefer people to use their regular modeler then convert the output using the ThinkFish styles tool. If you want to see what their internal tool can do take a look at Berkeley Systems new Java Joe screen saver.

Is the fish on the ThinkFish Web site a demonstration of their technology? How come there's no need to use a viewer to see it? Its a fake - an animated GIF to give the impression of seeing their fish character. Its about 17k as a seven stage animation. You need QD3D to see their output at the moment. Can people make their own styles? No, maybe one day, but that would be a high-priced developer tool at around the one or two thousand dollar mark. For now they are doing their best to be ubiquitous so look out for ThinkFish.


Next on was Mark Meadows and Lisa Goldman from Construct. Mark carries the titles VRML architect / animator / Webmaster, Lisa is President. Construct's been around for over 18 months now making VRML worlds - they got in from the earliest beginnings of VRML back in Ono Sendai days and made one of the first 3D worlds: a model of the Interactive Media Festival in LA. At the moment they are engaged in exploring 3D comics as a format because "a bit of abstraction helps things out a lot".

By using the conventions of a comic they have found ways to circumvent both the need for photorealism in characters and a way to progress a narrative story using VRML. Involving visitors to a site in a story means they will stay longer, have more commitment to the site, and hopefully want to return. Construct are working to integrate different media types into the comic format, starting with sound to replace text as a means of moving the story along. "VRML as a graphics protocol is still pretty weak" says Mark but he's done some amazing things with it none the less.

Crutch is a comicbook style story still in development by Construct (visit their Web site for the latest version: construct.net/projects/crutch). The story is about a man who starts off walking in the desert, uses a stick for a crutch then gradually has to replace parts of his body as they fall off under the strain of using the crutch. There is a Promethean part where a bird flies off with the man's liver and drops it at a gas station. The ending is still being worked on but the story loops back to a beginning from which it can run round again.

The artwork is developed using a mixture of natural media types, PhotoShop images, Alias models, Cosmo Worlds for rearranging the scene graph, and VRML hand coding. Mark snapshots the VRML scene then works on the image in PhotoShop to produce a frame of moving action surrounded by still shots of artistically enhanced angles.

The VRML scene is displayed using an animated camera viewpoint. Mark discussed some of the frustrations of working in VRML at the moment: the inconsistencies between the browsers, the different rendering engines ways of handling color, lack of control over the color palette, audio differences between browsers, frame rate variations depending on machines and window size, all the issues to do with external interface controls that the working groups are laboring over at the moment.

He also talked about how to take advantage of using tricks like sloppy vertex shading, jostling the polygons, and roughing things up to give an artistic look to the scene. He was definitely in agreement with ThinkFish when it came down to how to give a scene artistic merit as opposed to realism. But he's not a pure artist: he has a coder's tendency to hide things in the source code. If you want to find out more about the narrative go look at the comments in the VRML file. As the scenes are VRML you can choose to abandon the prestructured narrative flow and explore for yourself.

Mark showed us a novel way of visualizing the structure of the story: he has a VRML model of the storyline as separate rings floating in space, each segment of the ring being a scene in the story. Where the storylines of the man and the bird intersect the rings are linked by other 3D structures. The flow of time is showed by an arrow that rotates around the rings. This set up allows a connected semi-linear plot to be constructed with all the storylines starting and finishing in the same place.

The second demo, Dendrite, showed us another story, this time set in a world of half-man-half-fish creatures that travel to work along long tunnels that are some sort of allegory for the brain or a sewage system. Mark's aim is to preserve a sense of mystery even in a factory-like world. The sequences he showed, of his man-fish swimming down a long corridor to catch his transportation to work was certainly weird and atmospheric. There are other characters in the corridor world: krill that are food creatures and an enemy that is trying to destroy the system. This seems a story that will take more than one 'read' to understand. Mark told us "The need for a narrative line via text or audio is under experimental investigation right now." He believes we have to go through a gestation period, a learning curve for readers to understand the new narrative structures that are being developed.

Mark contrasted the innovative work Construct are doing with the more conventional 40's and 50's style cel animation done by Protozoa, famous for their Flook series on SGI's VRML site: vrml.sgi.com. Mark thinks it could be another couple of years before the market is ready for an industry based on original 3D content. In the meantime he says: 'I'm trying to see what we can do with this technology. VRML is exciting to use so it'll be in all our comics." Entirely open protocols mean anyone can see it - no-one has to download a special viewer - Construct is in favor of openness.

Coming next: On Tuesday 20th May at 6.30 for Schmoose, 7.30 for presentations. Where are the clients and what do they want? What motivates clients to use this medium? What challenges does VRML solve? Learn from two presenters who have a unique access to the client base and find out the lessons they have learned. Presenters are: Planet 9 Studios and TechniCom Computer Services.

The VRML SIG is held in Cafe Iris at SGI's offices in Mountain View CA. Food will be provided. Directions: Take the Shoreline Blvd. exit off Hwy. 101. Head East toward the shoreline amphitheater, At the fourth stoplight (Stierlin Court) turn right (one light after Charleston) and the fourth building on the right (building 20) is Cafe Iris. Cost: free to Software Forum members, $10 to non-members.

The VRML SIG is sponsored by Fujitsu Software Corporation through the Software Forum. To find out more contact Katherine Bretz at kbretz@fsc.fujitsu.com or see the Software Forum Website at: www.softwareforum.org, and VRMLPro at inquiry.com at: www.inquiry.com

Videos of the proceedings are available for $25 from Fem Energy Box 1176 Boulder Creek CA 95006, (408) 338-7228 or sunrize@cruzio.com The URLs for the presenting companies are:

Construct: www.construct.net

ThinkFish: www.thinkfish.com

Usually found online at inquiry.com as the VRMLPro, Sue writes regularly for Web publications on VRML and 3D graphics. Contact her at: vrmlpro@inquiry.com.

Back Contents Next
Have you tried out our new Discussion Area yet?

LinkExchange Member

Need sponsorship or advertiser information? E-mail adrian@aereal.com.

Have an idea for an article? Thought of a way we can improve VRMLSite? E-mail suggest@vrmlsite.com.

Please Tell a Friend about VRMLSite!

© 1997 Aereal, Inc. All rights reserved. VRMLSite is a trademark of Aereal Inc.