AI voice transcription and captioning platform Verbit has updated its Captivate ASR solution to enable it to identify specific features in automated captioning.
According to the company, the update is an “industry first”, with the identification feature enabling viewers to see captions and subtitles that clearly identify the speaker versus generic chevrons (>>) or dashes traditionally used to indicate speaker transitions.
For example: >> JONATHAN WILLIAMS: Let’s look at some of today’s top stories in the news.
The company’s Global Prep Team captures voice profiles, or “voice signatures” from designated speakers, such as presenters, reporters or sportscasters, before a programme goes to air. These signatures are labelled, added to Verbit’s trained acoustic and language models and activated during live broadcasts to accurately and clearly tag who’s speaking in real time, said the company.
The speaker ID feature is available across news, weather and live sports.
“Live ASR caption viewers deserve the same clarity and context that human captioning has long provided,” said Verbit general manager Doug Karlovits. “Our new speaker identification solution leverages the most advanced and innovative speaker models—far surpassing traditional ASR outputs—to achieve the highest accuracy for speaker IDs.
“We work with customers to determine which speakers they want to identify,” added Karlovits. “And as with all our services, we offer a range of customisation options for speaker IDs and can tailor formatting and styles to specific customer requests and preferences.”
