Avatars behaving badly

Two pieces of déjà vu hit me today, reminding me how ideas get constantly recycled, and how long it can take for some technologies to find a viable application. (Remember "push" technology of ten years ago? Now it's back as RSS). 

First, out of the US election hullabaloo came JibJab media’s Bush-Kerry cartoon, taking the Internet by storm, and racking up 2.5 million hits a day. It’s silly, sophomoric, satirical, and very cool. Second, out of the UK comes the news that customer-friendly ATMs that actually use your name on-screen are finally going to make an appearance in that traditionally techno-skeptic land across the pond.

The connection? In the mid 1990’s I was running a company in London that was pushing the envelope in video-realistic 3D avatar technology. With just two photographs of a person we could create a life-like 3D virtual character that worked on an ordinary PC or over the Internet, complete with facial expressions, moving eyes, blinking, and rather good automatic lip-synching to a voice track.

My goal was to use these avatars in training games, simulations, and online communities (we were working with Worlds Inc. in San Francisco), but in order to pay for their development I was looking for other applications. I tried pitching the talking head to NCR for use in ATMs – it would pick up a text feed, convert it to voice, and synchronize the lips and facial expressions of the ATM character, and address the customer by name. But, partly because it was so real, the effect was a little creepy and developers were not interested. Eight years later ATMs are still text-based, but are finally acknowledging the customer by name.

What about Bush-Kerry? We exhibited at the E3 video games expo in 1996, and to promote the technology we put together a set of give-away screensavers that satirized the candidates in the then-current elections. “Raucous Caucus” featured 3D avatars of Clinton and Dole debating and performing in suitably scandalous sophomoric fashion. Had the Internet been bigger at the time, it would have been huge. Or maybe not – our technology was generations ahead of the Bush-Kerry cartoon du jour, but our content was nowhere near as clever. Talent trounces technology, and substance beats style every time.

One thing that became really obvious as a result of all of the avatar work, was that believability is less about appearance than about behavior. In simulations, a very large portion of your budget can go to replicating look, when getting feel right pays much bigger dividends. People will forgive appearances if the behavior is realistic; but no amount of visual gloss can make up for unconvincing behavior.