ΒιΆΉΤΌΕΔ

Research & Development

Abstract

As broadcaster’s web sites become more media rich it would be prohibitively expensive to manually caption all of the videos provided. However, many of these videos have been clipped from broadcast television and would have been captioned at the point of broadcast.

The recent FCC ruling requires all broadcasters to provide closed captions for all β€˜straight lift’ video clips that have been broadcast on television from January 2016. From January 2017 captions will be required for β€˜Montages’ which consist of multiple clips, and the requirements to caption clips from live or near-live television will apply from July 2017.

This paper presents a method of automatically finding a match for a video clip from within a set of off-air television recordings. It then shows how the required set of captions can be autonomously identified, retimed and reformatted for use with IP-delivery. It also shows how captions can be retrieved for each sub-clip within a montage and combined to create a set of captions. Finally it describes how, with a modest amount of human intervention, live captions can be corrected for errors and timing to provide improved captions for video clips presented on the web.

This document is based on one originally published at NAB 2015. Because it was written for an American audience it uses the word β€œcaptions” throughout in place of β€œsubtitles”.

Topics