Imagine addressing your Amazon Echo speaker by the bedside: “Alexa, can Grandma finish reading me The Wizard of Oz?” And Grandma’s voice fills the room, even though she passed away a couple of years ago.
Amazon has demoed the first glimpse into Alexa’s newest feature, which is still in development. The voice assistant can replicate people’s voices from short audio clips, as short as one minute. “This required an invention where we had to learn to produce a high-quality voice with less than a minute of recording versus hours of recording in a studio. The way we made it happen is by framing the problem as a voice conversion task and not a speech generation task,” said Rohit Prasad, senior vice-president and head scientist of Alexa artificial intelligence during the company’s conference at Las Vegas.
Though a timescale hasn’t been provided for the launch of the feature, the underlying technology has existed for several years. Though the application is being discussed on social media and some are calling it “creepy”, AI voice mimicry is not new. We have already seen “audio deepfakes” used in industries like podcasting, film and TV, and video games.
Last year, Anthony Bourdain’s voice was recreated with AI for a small part in the film Roadrunner. Morgan Neville, the director of the documentary had said he used AI to generate Bourdain reading from letters he’d never spoken aloud. A couple of years ago, Joshua Barbeau trained a version of the GPT-3 chatbot on conversation logs with his late fiancee Jessica, who had died eight years earlier. And then there is Eugenia Kuyda who had built a chatbot out of her partner Roman Mazurenko’s old text messages.
Some audio recording suites offer users the flexibility to replicate individual voices from their recordings to edit parts that had gone wrong. But there is always the risk of blurring the lines between what is human and what is mechanical.
Though Prasad didn’t address the issue of the technology being misused, he did say that the ability to mimic voices was a product of “unquestionably living in the golden era of AI, where our dreams and science fiction are becoming a reality”.
Prasad said the feature may enable customers to have “lasting personal relationships” with the deceased.
