Cepstral Adds Open Source MRCP Library to Telephony Industry

Cepstral, a text-to-speech voice provider, is taking steps in propelling the developer community to produce more effective, meaningful speech technology applications.

The company's funding of an open-source Media Resource Control Protocol (MRCP) library comes as the first time that  MRCP sample code has been available. In the past, companies spent upwards of a year developing MRCP tailored to their businesses; they had no choice, as the standards protocol remains widely adopted among many of the largest telephony platforms and speech technology vendors. Craig Campbell, Cepstral CEO, says the library's release is all part of a larger development process.

"For the first time, there's a common base of code from which people might be able to write to it," Campbell states. "If we can get that common reference point, we might be able to write small wrappers that can achieve that dream of portability and interoperability, rather than spending years writing your own MRCP stack to communicate with a speech engine."

MRCP allows speech services to distribute speech services (speech recognition, speech synthesis, etc) to clients - while the server runs on one computer, clients send MRCP messages to the server over a network. Similar to Internet protocols, MRCP code contains information for the server in sections: the type of information and response codes; and a body containing the message's details. Re-inventing entire protocol layers, says Campbell, kept companies from speedier developments, and from adopting newer versions of  MRCP.

"I know there are groups out there who have implemented Version 1, and they're in a wait-and-see mentality with version two," Campbell adds. "The promise of MRCP is portability, so like many things it's a work in progress. I think there's a lot of companies able to use MRCP who can now take the next step toward version two in particular, if they need that."

And, because the MRCP stack is open source, Campbell also stresses the importance of peer edits, comments, and suggestions. He says he hopes those who gain from the site's resources will also give back in some form.

"One of my hopes is that this gets traction, so people not only use it but contribute to it," Campbell says. "People could see things in there that they might want to make better, or see something they did that would make it better; that sort of momentum is what would make MRCP a real baseline if it gets the support of the community."

SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues