New Software Uses Phonetics To Search for Video Clips

Television coverage of major international sports events like the World Cup and the Olympics generates huge amounts of video footage, which is often difficult or time-consuming to search through and find. A possible solution to this problem could be a new software package that recognises phonetics in speech on the audio tracks of clips.

“Get” for Apple Final Cut Pro (FCP) was developed by UK company AV3 Software and is currently in beta test at 40 facilities worldwide ahead of its official launch in July.

AV3 was founded by Graham Sharp and Peter Ramsay, who had previously worked at Avid in the UK and realised that increased amounts of footage called for an efficient method of finding the right clip. “Graham and I saw that a lot of time was spent by people in the broadcast industry searching for material,” says Ramsay, who heads AV3 as managing director. “Shooting ratios have increased: for any given half-hour programme, there can be 15 hours of source material, so how do you find the shot you want?”

AV3 formed a partnership with American development company Nexidia, which produces phonetic-search technologies for the IT sector. Using Nexidia’s core program, AV3 has produced a system tailored to the broadcast market. “It’s not a ‘new’ new product,” says Ramsay, “but it is a different iteration of Nexidia’s system.”

“Get” for FCP, in what is considered Version 1 of the program, works only if the Apple system is connected to a drive. There are plans to extend it to other editing workstations.

The phonetics program handles seven languages: Dutch, English (in its North American, British, and Australian variants), French, German, and Spanish. Nexidia recorded samples of 100 speakers of each language to provide enough variations in how words are pronounced.

Ramsay says he wanted a good cross-section of users for the beta-test period, including broadcasters and facilities involved in sports and news broadcasting, documentary making, and corporate production.

Sports TV in particular involves searches through a lot of source material in assembling packages and programmes. To use “Get” for FCP, a phrase — the name of England footballer Wayne Rooney, for example  — is typed into the program window, and a search is made of audio tracks on all video clips held on a server or computer drive.

Ramsay acknowledges that the system can sometimes pick up on the wrong thing but says the tolerance parameters can be changed to make it more or less accurate.

A further feature of the system is that, when new drives are added or new footage loaded on to existing drives, “Get” for FCP will search to find any new mentions of a programmed phrase, such as “Wayne Rooney.”

“Get” for FCP won the Black Diamond Award at NAB this year, where AV3 demonstrated the program and put out a call for beta testers. The program is currently being tested by users in Europe, the U.S., and Australia.