Architecture Converts Voice Into Streaming

Welcome Guest. | Log In| Register | Membership Benefits

July 24, 1998 (12:26 PM EDT)

A Colorado Springs, Colo., start-up has defined a server architecture capable of converting analog voice messages from the circuit-switched phone network into streaming, compressed RealAudio files.

The iTalk Server architecture from TellSoft Technologies, a company formed by telecommunications consultants formerly with MCI, supports near-instantaneous conversion to ASM or HTML assembly-language code, making it possible to add a recorded voice message to a website seconds after the words are spoken.

Even though TellSoft has been in business for only a matter of months, it is already listed as a primary partner in RealNetworks' Web list of G2 development-environment beta users.

Originally, the company was pursuing ISPs, website developers, and small-vertical-Web users such as real estate agents as initial customers of iTalk. But TellSoft reported a high degree of interest among European private branch exchange and call-distribution manufacturers, suggesting the server architecture may be more widely applicable than anticipated.

The iTalk system is based on an architecture TellSoft calls iCT ("ice tea"), for Internet-enabled computer telephony. Company president Shaun McNerney emphasized it is by no means limited to audio files. The system can be adapted to a variety of multimedia data types and codec hardware-circuit algorithms, and a mix of embedded message types, McNerney said. At its heart is a set of conversion facilities based on C++, for which TellSoft has sought numerous patents.

"We keep a continuing eye on the video market, but the use of streaming video on websites is nowhere near the acceptance levels of audio yet," McNerney said. "But if we want to add video, if we want to move to new compression algorithms, if we want to extend hardware support, it's a matter of simple add-ons, not a redesign of the entire system."

Developers with direct control of an iTalk-enabled website can set quality-of-service parameters for different users. They might, for example, allow a company's executives access to high-bit-rate voice compression, while confining external users or lower-level employees to more bandwidth-efficient codecs.

For a user dialing in to an iTalk site, however, all such parameters are transparent. The system operates through simple push-button commands akin to voice mail.

Audio messages converted into compressed files are immediately available for playback, either from a website or from a dial-in phone system. This could allow iTalk to interface in unique ways to systems using simpler Web-access languages such as the Handheld Device Markup Language (from Unwired Planet, as well as to voice-response systems from General Magic and other companies. TellSoft has opened discussions with a number of companies in the embedded voice-response field, though no deals are imminent.

"The real trick is keeping up with all the customers who come back with unique applications for embedding voice," said TellSoft vice president Paul Carver. He said the company would like to be able to release a software developer's kit to enable third parties to commercialize some possible voice applications. But for the time being, TellSoft is simply scrambling to keep up with beta users' development of proprietary applications.

While iTalk is based on proprietary software, TellSoft includes Dialogic time-division multiplexing telephony boards among its deliverables. Creation of an iTalk system requires defining a Windows NT server as an iTalk server, and adding Dialogic boards to handle voice services.

In most applications, the server will be separated from both the Web server and the RealNetworks Real Server, though server functions can be combined in some embedded applications.

The system was ported to NT initially because of market size, but TellSoft is considering ports to Solaris, Linux, and other operating systems. In an NT environment, iTalk does not require more than a 486 host, though a 200-MHz Pentium is recommended. No special MMX extensions are required.

The server establishes real-time links between the Dialogic programming interfaces and the RealAudio software codecs, while the software manages all multithreading of the audio files. A special "Publish" facility takes care of real-time conversion to HTML, ASM, or other formats.

TellSoft will be pricing smaller application bundles at $2,995 for servers supporting two voice lines, and $4,995 for four-line systems. The company is working with RealNetworks and other larger companies to have its iTalk logo recognized as an audio-enabled hyperlink.

McNerny said he's keeping his eye on companies offering specialized voice-mail applications, as well as the potential of alternative concepts for file conversion, including Java applications for telephony. So far, though, TellSoft executives said they have not found another architecture capable of scaling to the large number of users found in national circuit-switched phone networks.