The BBC Sounds app has launched a trial using generative AI to create subtitles and transcripts for a range of audio programs on the platform.
The broadcaster said it is using an AI speech recognition tool called Whisper, developed by ChatGPT developer OpenAI and made available as open source, for the new feature. The new feature is being tested publicly on the web and Android versions of the BBC Sounds app, with support for Apple’s iOS to follow in the next few weeks.
It said the tool would be used to “quickly produce a high-quality transcript” of the programme audio before it was reviewed and edited where necessary by BBC editors. A final transcript containing the audio would then be uploaded to the BBC Sounds app.
Aniruddh Dimri, head of product at BBC Sounds, said: “A crucial part of the BBC’s mission is that everyone across the UK feels the BBC is there for them. Sometimes that’s about making sure the content reflects our different backgrounds, interests and identities, but it’s also about making sure everyone can access our content.
“BBC Sounds, for example, currently produces around 27,000 hours of content a month – but much of this is difficult to access for the approximately 18 million people in the UK who are deaf, have hearing loss or tinnitus. We looked at ways to add subtitles so people can follow the audio using text.
“Doing this manually would be time-consuming and prohibitively expensive.
“However, as we test new technologies and explore how we can work with generative AI tools and use them to benefit our audiences, we are exploring whether AI can help us add high-quality and accurate subtitles to our audio content.”
As part of the trial, the AI-supported transcripts will initially be used in programs such as “In Touch,” “Access All,” “Profile,” “Sporting Witness” and “Economics with Subtitles.” Over the course of the three-month trial, the BBC said it intends to use the technology in other programs as well.
“After three months, we will review the progress made in the test, how well the tools work and whether they represent a cost-effective way to make transcripts available in Sounds,” said Mr Dimri.
“Following this review, we will decide whether or not to proceed and, if successful, whether to extend it to more of our content on BBC Sounds and potentially to our archive as well.”
Since the release of ChatGPT, generative AI tools have become increasingly common over the past 18 months as technology companies and other firms have sought to capitalize on the hype surrounding the technology and provide new tools to attract more consumers to their products and services.
