The BBC is looking to make its podcasts and other content delivered via BBC Sounds more accessible to people who are deaf or hard of hearing issues. To achieve, the broadcaster is turning to generative AI technology.
Beginning on Aug. 29, 2024, the BBC is bringing subtitling to the BBC Sounds app for a three-month trial. According to a blog post from Aniruddh Dimri, head of product for BBC Sounds, manually captioning the 27,000 hours of content produced monthly for BBC Sounds would be both time consuming and “prohibitively expensive.”
Instead, Dimri wrote, “we have been looking at whether AI can help us add high quality, accurate subtitles to our audio content.”
BBC is using Whisper AI, an automatic speech recognition system. According to Open AI, Whisper has been trained to understand a wide variety of accents and technical. The programs are run through Whisper to generate a transcript that is then reviewed by an editor before it is added to the audio program, according to Dimri.
The trial is currently only in the Android version of the BBC Sounds app and on its web-based player. An update for iOS is expected “in the coming weeks.” Programs in the trial are “In Touch,” “Access All,” “Profile,” “Sporting Witness,” and, appropriately enough, “Economics with Subtitles.” On subtitled programs, users can click a speech balloon button to access the subtitles.
Currently, Dimri, said the trial will run for three months and then BBC will review to determine its next steps.
In its 2023/24 Annual Report, the BBC reported that nearly 100% of its broadcast hours on BBC Two, BBC Three, BBC Four, CBBC and CBeebies included fully subtitled content, as well as 99.9% of the broadcast hours on BBC One, BBC Scotland, and BBC News, and 50% of broadcast hours on BBC Parliament. For BBC iPlayer, 89.6% of content is subtitled.
