NEWS

WIRES

OTHER

Cepstral Breaks the 2 MB Voice Barrier

Tiny Text-to-Speech Voices Poised to Make Big Waves

Cepstral LLC announced a new software release capable of producing fully general, unit selection Text-to-Speech (TTS) voices that are under two megabytes in size on the eve of the Consumer Electronics Show, CES, in Las Vegas.

Cepstral's new Swift 3.1 engine is the first high-quality human sounding voice based on unit selection technology to reach this threshold. "This puts high quality Text-to-Speech within reach for consumer electronics manufacturers enabling them to place features onboard such as talking e-mail, talking e-books, spoken reminders, or talking directions," said CEO and co-founder Kevin Lenzo.

By scaling high-quality voices down to 2 megabytes, they can potentially fit inside today's compact consumer electronics. This is important because the features afforded by TTS solve several real problems. For instance, there are safety concerns over portable devices being used while operating a vehicle. Text-to-Speech facilitates connectivity by offering a hands-free, eyes-free interface which in turn increases the utility and value of such devices.

Devices that help people manage dynamic data are especially well suited for TTS. A synthetic voice offers unconstrained content delivery so whether the device is an email tool, navigation tool, learning tool, or bio-metric tool, Cepstral's TTS engine can speak the results.

While tiny in size, these software voices can say anything. Currently, they are available only as male and female US English voices, but will soon be available in a half-dozen languages. Cepstral's voices are platform independent, running on Windows, WinCE, Linux, and Palm powered devices. The voices are scalable by design and can be tailored to meet specific size requirements anywhere from 2 MB and higher. "The bottom line is Swift 3.1 is the smallest high-quality, unit selection TTS engine on the market. Even at that, the Swift engine outperforms competing products that are hundreds of times its size in terms of platform independence, speed, and pronunciation accuracy," said Mr. Lenzo.

Cepstral was founded in 2000 by Dr. Alan W. Black and Kevin Lenzo, two renowned speech synthesis scientists from Carnegie Mellon University. The company specializes in producing high-quality, small-footprint TTS for the embedded electronics markets.

For more information about Text-to-Speech, or to schedule an interview with Kevin Lenzo, please call Craig Campbell at 412/432-0400 or craig@cepstral.com