Is it real, or is it LifeFX?

A small company in suburban Boston has achieved a goal that has so far eluded even the mightiest of Hollywood special effects machines – it has made an image of an old woman smile convincingly. Digitally.

Made not of flesh and blood but instead of bits and bytes, she is a digital creation of LifeFX Inc. in Newton, Mass. What is most remarkable about her is that she moves and speaks in a lifelike way, even when viewed so close you can count her pores.

The goal of creating a lifelike human animation has been elusive because we move so many muscles as we speak or even listen. Brows knit, eyes roll, lips curl up in a smile then purse, cheeks puff out. The skin is smooth or wrinkled, it has large or small pores, it draws taut or crinkles at eye corners. And consciously or not, we notice every little detail.

LifeFX’s old woman isn’t perfect – her eyebrows aren’t symmetrical, her smile is a bit crooked – and that may be the secret of the company’s success.

“It is its imperfections that makes something look real,” said Louis Latham, an analyst at Gartner Group Inc. in Stamford, Conn.

Video game developers first made a perfect computer-generated spaceship, but “its perfectly smooth, shiny surface looked cheap and unreal,” Latham said. “They found that by adding some dings and scuffs, and making it less perfect, they made it look more real.”

Initially, LifeFX had its eye on the entertainment field, said Lucie Salhany, co-president and CEO of LifeFX. “But as the Internet grew, we saw a growing need for a better human-to-machine interface.”

The company also has strong roots in medical research. Early models were built to increase understanding of human physiology and for use in telemedicine and distance learning.

LifeFX builds the human head from the inside out, first describing eyeballs and tongue and how they move. Then the teeth and jaws. It builds a skull and attaches each muscle, then covers its flawed human with skin. Each bit of flesh and bone, each ligament and tendon and how it moves is digitally described.

As the digital human blinks, the application draws on this 3D database to describe how underlying muscles cause the eyelid to fold back into itself, drawing the skin with it.

By altering the database to reflect the changes that occur with age – a decline in bone and muscle mass, joint flexibility, and muscle and skin elasticity – and how those alterations in characteristics affect the way the body moves, the application can digitally age a young woman. By adding a text-to-speech engine, LifeFX’s digital creatures speak in a lifelike way.

LifeFX’s first commercially available product, Facemail, was released last month and is available free at the company’s Web site. Facemail offers users a choice of five different 3D digital stand-ins to deliver their e-mail messages.

After downloading the free Facemail player, users select a stand-in, who will deliver their talking e-mail. Users type their message, a text-to-speech engine converts it to audio and synchs it with facial movements of the stand-in, and the audio e-mail is sent. To see and hear the e-mail, recipients also must download the Facemail player.

By adding any of a half-dozen “emoticons,” users can cause their stand-ins to make an appropriate facial expression at a particular spot in the message. For example, instructing the image to wink as it said, “You want me to work the weekend? Oh, sure, sign me up for that,” would presumably clear up any confusion over the likelihood that you will show up for work Saturday morning.

In simplifying the software to run over low-bandwidth connections, some verisimilitude is sacrificed. If the stand-ins are less convincing, the audio and video are at least in sync.

Later this year, Facemail will let users speak into a microphone and have their stand-in deliver the message.

By the end of next year, users will be able to send LifeFX a digital or analogue photo of themselves and get a stand-in of themselves to use in e-mail, chat rooms and instant messaging.

Production: Please make Sidebar:

Novelty or new business tool?

By Sami Lais

Computerworld (U.S.)

The novelty of having a digital human read your e-mail aloud will attract Facemail’s first users but its effectiveness as a business tool will retain users, said spokespeople for LifeFX Inc.

The company last month released its first commercial product, Facemail, which lets users select one of five stand-ins, as LifeFX calls its digital avatars, to deliver their e-mail.

LifeFX is negotiating with two Fortune 100 companies to license the 3D talking heads for use on their Web sites, said Lucie Salhany, CEO and co-president of LifeFX. “Shopping-cart abandonment is a big problem,” she said. “Suppose there’s someone to help you through the process, or even to say, ‘Are you sure you don’t want to buy this?’ “

Perhaps, said Jackie Fenn, an analyst at Gartner Group. “But if you’re at work and you’re busily buying a sweater on-line, you may not want someone on your screen announcing it.”

And although Facemail’s inflection and visual cues may cut down on misunderstandings engendered by e-mail, as LifeFX claims, “it may prove disruptive in an office full of cubicles,” according to Fenn.

But e-mail is only the beginning, said Mark Sagar, vice-president of research and development at LifeFX. By linking the visual technology to artificial intelligence and an English discourse engine, a lifelike digital human can interact with users.

“Think of a HAL-like creature,” Sagar said, referring to the computer in Stanley Kubrick’s film 2001: A Space Odyssey.

“It’s a natural for telemarketing, call centres, help desks, any repetitive tasks,” Sagar said. It can also cut costs. “The average customer service call costs $US25 to US$33 per incident. Think of the savings if you can automate that.”

Much of what its avatars can do can also be done “in text, but it’s cold,” Salhany said.

“The biggest difference [the technology offers] is in adding sound,” Fenn said. But in “on-line retail, for a help desk, it could work. People react differently when a machine is humanized. Look at speech recognition – people start saying please and thank you, and they know they’re talking to a machine.”