Upon careful evaluation of ENCO’s enCaption4 automated closed captioning system at four broadcast trade shows, Colombian TV channel Citytv opted to put the challenges of manual captioning to rest. The TV station serving the capital city region of Bogota has since increased speed and accuracy while reducing captioning costs by 75 percent.
Citytv, owned by media company Casa Editorial El Tiempo, licenses the Citytv brand from Canadian communications company Rogers Media. The station, which broadcasts on UHF channel 21, follows a mandate that requires Colombian over-the-air TV stations to provide closed captions in 75 percent of their programs between 6am and midnight. Citytv captions its daily news program, Citynoticias, along with other popular shows such as Arriba Bogotá and Bravissimo.
The speed by which enCaption4 generates closed captions, especially compared to the manual “listen and talk” speech recognition system that it replaced, was a key factor in Citytv’s decision. Building on ENCO’s patented automated captioning approach, enCaption4 combines machine learning with a neural-network speech-to-text engine to deliver exceptional accuracy with extremely low latency.
Gomez notes that enCaption4 is always ready to caption, and can operate 24/7 in a completely automated manner. While the system can be deployed on premises or in the cloud, Citytv preferred to install it locally in its broadcast master control room. There, it’s configured to receive a program video signal with embedded audio. The captions generated by enCaption4 travel over IP to an Evertz encoder, which embeds them into the live broadcast video.
Citytv also interfaces enCaption with its MediaCentral (iNews) newsroom computer system (NRCS) to familiarize it with specific words that it will interpret into captions. Using MOS standard protocol, enCaption4 can access news scripts and rundowns from any NRCS and learn correct vocabulary, spellings, names, places, and other special content that will be mentioned in the local newscasts. It retains these terms in its ever-expanding internal dictionary, becoming more valuable to the user over time.
Continual product development has also resulted in enCaption4’s ability to distinguish between multiple people speaking. This new multi-speaker distinction functionality leverages AI to detect changes between speakers even within a single mixed feed. enCaption4 today supports more than 30 languages with accurate spelling, capitalization, and punctuation to strengthen caption quality.
