The following article is from the Pittsburgh Tribune-Review on Wednesday, April 8, 2009.
South Side firm develops program that reads text in your voice
Fingerprints and DNA might cement someone's identity with police or scientists.
"But our voice is probably the most important identity we have," said Bob Wilderman, 62, a Philadelphia-area resident who suffers from ALS, or Lou Gehrig's Disease.
Before the illness robs him of his voice, Wilderman is working with South Side-based Cepstral to record it.
"Then I'll be able to communicate every day in phone conversations, on my ham radio or on the computer," said Wilderman, whom doctors diagnosed about two years ago.
Cepstral has developed technology that translates written text to speech. Unlike existing, similar software, Cepstral has created a method that allows customers to choose a voice. It could be one's own voice — as in Wilderman's case — or a voice of a certain age and gender or preferred geographical accent.
Translating languages, enabling the physically disabled to connect with others and teaching with computer tutorials are among the company's text-to-speech applications, CEO Craig Campbell said.
"But we're not focusing just on the utility value," he said. "We're also emphasizing the subjective value — people like what they like."
That means if a teenage boy wants his "World of Warcraft" character to sound older and more ominous, he'll have that choice. If an airline executive wants the company Web site voice to be calming, soothing and female, that option will be available.
Cepstral employees build voices by recording people speaking up to 4,000 words, phrases and passages. Another approach is to extract voices from the public domain on the Internet, said Patrick Dexter, Cepstral's director of business development.
Cepstral's breakthrough is particularly notable because computers could influence human emotions, said Lonnie Benson, 52, CEO of Humanity Interactive in Seattle.
"Asking, 'Will computers ever be as smart as people?' is the wrong question," said Benson, whose company uses Cepstral's technology to make its Internet-based avatars more lifelike. "The right questions are, 'Will computers make you happy? Will they make you smile?' That's where we are right now."
Inanimate objects talking to people soon will be the norm, said Alan Black, who founded Cepstral in 2001 before returning full-time to his position as a Carnegie Mellon University computer science professor. Pill bottles will announce what medication they contain, he predicts. Elevators will ask, "What floor?" Office doors will greet visitors like receptionists do now.
"Ten years from now, we'll wonder why things don't talk to us," Black said. "It'll be like checking into a hotel and learning they don't have Internet. We'll be like, 'What do you mean?' It'll be weird."
The same goes for interacting with people in 3-D Internet communities such as IMVU.com, which boasts 35 million members, said Jeff Titterton, vice president of marketing at the Palo Alto, Calif.-based company.
"When people first started to meet through online dating, many thought that was bizarre," Titterton said. "By the early 2000s, it became an absolutely acceptable mode of communication. Our virtual world is the next evolution beyond Facebook and MySpace."
Cepstral's text-to-speech technology allows users to be more creative and expressive, Titterton said.
"We can extend the illusion that you're talking to the person in real life," he said. "That's neat."
Mike Cronin can be reached at mcronin@tribweb.com or 412-320-7884.
