MediaSite Debuts World’s First Speech Recognition Products Designed for Media Applications
MediaSpeak Solo and MediaSpeak Chorus Designed for Video and Film Production

Las Vegas, NV—April 6, 1998—MediaSite, Inc., the leading developer and supplier of integrated solutions for constructing and accessing network-based, searchable digital media libraries, today announced the debut of MediaSpeak Solo (TM) and MediaSpeak Chorus (TM), the world’s first family of speech recognition software designed specifically for media applications.

With over 10 million hours of existing video and the constant creation of even more content, news and film organizations, government agencies, and corporations require a sophisticated solution to manage these video assets. Speech recognition is a critical part of that process, which is why MediaSitehas developed the MediaSpeak family of speech recognition software. With MediaSpeak software, organizations can automatically recognize the audio portion of live and archived video, and catalogers and producers can use spoken annotations to quickly describe more detailed information about video.

MediaSpeak Solo and Chorus both convert speech to text, but each product features a distinct type of speech recognition technology for the unique needs of different media applications. MediaSpeak Solo (a speaker dependent product) is optimized for a single user speaking into a microphone, who trains the software to recognize their speech pattern. With MediaSpeak Solo, loggers, producers, and librarians can now use speech recognition to describe video, such as news, that was typically annotated via typed text. MediaSpeak Chorus (a speaker independent product) is optimized for multiple unknown voices and can recognize audio sourced from VTRs, live video feeds, microphones, or digital format. This makes MediaSpeak Chorus ideal for automatically processing the multitude of different speakers found in news, film, documentaries, training, and conference panel sessions.

MediaSpeak Solo and Chorus are available as licensed software for developers or as an integrated part of ISLIP’s industry-leading MediaSite Digital Video Library System(TM), a complete software solution for converting live and existing analog video to a network-based, searchable digital format.

"With speech recognition now integrated into video asset management systems, a host of applications can now take advantage of this revolutionary technology, especially the ability for a system to recognize multiple, untrained speakers," said Mark Juliano, president and CEO, MediaSite, Inc. "As we head into the age of digital media, organizations now can automatically capture the full experience of news, sports, events, and conferences and preserve this information via computer. Speech recognition is a core tool to make this content easily searched and re-used in ways not possible today."

MediaSpeak Solo
MediaSpeak Solo single-speaker speech recognition software is ideally suited for spoken annotation in real-time logging environments. For example, newsrooms can receive up to 50 live incoming feeds which are annotated today via typed text by newsroom loggers. With MediaSpeak Solo, loggers can now describe video content using spoken or typed segment annotations. Since the average person speaks up to 3-5 times faster than they type, the use of MediaSpeak Solo dramatically reduces annotation time, while enabling loggers to capture more detailed and comprehensive video segment information.

MediaSpeak Chorus
MediaSiteoffers MediaSpeak Chorus multi-speaker speech recognition software in response to the need to automatically recognize the untold hours of videotape, film, conferences, panel discussions, training tapes, and news documentaries that would be unthinkable to process manually. In these applications, multiple people are talking who may not be identified and are certainly not available to train a single-user speech recognition product, such as MediaSpeak Solo.

MediaSpeak Chorus multiple-input technology recognizes audio sourced from a VTR, microphone, live feed or digital file format, while single-speaker products can generally only recognize input from a microphone. Chorus automatically adjusts to audio signal fluctuations and can be automatically trained to "learn" the terminology of specific industries, via custom dictionaries. Chorus can also learn word sets, such as "nightly news," and gender-specific acoustics models. This powerful combination of features allows organizations to automatically capture audio information and integrate it, along with video, into a digital video library.

Seamless Integration for End-to-End Digital Video Library Management and Re-Use
MediaSpeak technology is also embedded in the existing MediaSiteSite Digital Library System, the only end-to-end suite of software and services for constructing and managing digital video libraries. MediaSite Logger(TM) for real-time applications, MediaSite Builder(TM) for full-content indexing, and MediaSite Finder(TM) for sophisticated search and retrieval enable organizations to finally unlock the value of their video content by converting their video assets to an easy-to-use digital medium. Unlike other media asset management systems that are designed only for production and rely on simple annotations or titles for search and retrieval, users can search on the actual transcript of the video (derived from speech recognition and language understanding), and on select images such as faces (derived from image understanding), to pinpoint information or assets within the video.

System Requirements
MediaSpeak Solo and MediaSpeak Chorus operate in Windows 95/NT environments. MediaSpeak Solo requires a PC with an Intel Pentium 200 MHz MMX processor or higher, 64 MB RAM, and a Creative Labs Sound Blaster 16 or compatible audio board. MediaSpeak Chorus requires a PC with an Intel 266 MHz Pentium II processor, 128 MB RAM and 16-bit audio board.

Pricing and Availability
MediaSpeak Solo and MediaSpeak Chorus are available as licensed software for developers. Please contact MediaSite for specific pricing information. For end-users, MediaSpeak Solo is available as part of MediaSite Logger real-time cataloging system, available now and priced from $14,900. MediaSpeak Chorus is available to end-users as part of MediaSite Builder for full-content archives, available now and priced from $59,900.

MediaSite
MediaSite, Inc. is the leading developer and supplier of integrated solutions for constructing and using network-based, searchable digital media libraries. MediaSite offers the MediaSite Digital Library System, a complete set of software and services for developing digital audio and video libraries, including MediaSite Logger, MediaSite Builder, and MediaSite Finder. The unique feature of MediaSite’s technical approach is the integrated application of speech recognition, language understanding, and image understanding technologies based on software components of Carnegie Mellon University’s Informedia Digital Video Library Project. For more information, visit www.mediasite.com.