Editing Closed captioning (section)

==== Real-time captioning ====
Real-time captioning, a process for captioning live broadcasts, was developed by the [[National Captioning Institute]] in 1982.<ref name="caphist" /> As developed in 1992, real-time captioning used [[stenotype]] operators who are able to type at speeds of up to 375 words per minute provide captions for live television programs, allowing the viewer to see the captions within two to three seconds of the words being spoken.

Improvements in [[speech recognition]] technology mean that live captioning may be fully or partially automated. [[BBC Sport]] broadcasts use a "respeaker": a trained human who repeats the running commentary (with careful enunciation and some simplification and [[markup language|markup]]) for input to the automated text generation system. This is generally reliable, though errors are not unknown.<ref>{{cite web|url=https://www.bbc.com/news/uk-england-tyne-41473443|title=Match of the Day 2: Newcastle subtitle error leaves BBC red-faced|date=2 October 2017|work=[[BBC Online]]|access-date=2 October 2017}}</ref>

In the 1980s, [[DARPA]] sponsored a number of projects aimed at developing automatic speech recognition software. Much of this work was done by researchers at Carnegie Mellon University. In the 1990s, this program included a novel focus of using this technology for news transcription purposes.<ref>{{cite journal|pages=191–192|title=Speech Recognition by Machine: A Review|volume=6|number=3|year=2009|journal=International Journal of Computer Science and Information Security|arxiv=1001.2267 |last1=Anusuya |first1=M. A. |last2=Katti |first2=S. K. }}</ref> Later developments have yielded live, real-time AI-based captioning generating systems.<ref>{{cite news|url=https://www.theverge.com/2025/1/9/24339817/vlc-player-automatic-ai-subtitling-translation|title=VLC player demos real-time AI subtitling for videos}}</ref>