Today at the annual re:Mars conference in Las Vegas, Amazon SVP and Alexa Chief Scientist Rohit Prasad announced a number of new and upcoming features for the company’s smart assistant. Most dizzying of all was a potential new feature that could synthesize short audio clips into a longer speech.
In a script presented at the event, the voice of a deceased loved one (grandmother in this case) is used to read a bedtime story to a grandson. Prasad points out that by using the new technology, the company can achieve very impressive sound output using just one minute of speech.
“It required invention when we had to learn how to reproduce a high-quality voice in less than a minute of recording, as opposed to hours of recording in the studio,” notes the executive director. “We achieved this by framing the problem as a voice conversion task rather than a speech generation path. We doubtfully live in a golden age of AI, where our dreams and science are.”
Details are scarce at the moment. There’s no time frame or additional details, but at least it’s the kind of news that’s likely to spark a comprehensive scrutiny of potential applications beyond something as trite or even touching as a child’s reading. Wizard of Oz.
Credit: techcrunch.com /