During his re:Mars conference dedicated to artificial intelligence, Amazon presented a new feature of Alexa, allowing to imitate the voice of any human from a simple audio recording. According to the researchers, this project promises to be ” a speech conversion task and not as a speech generation path “. Based on this innovation, the American company hopes to bring people closer to their loved ones by voice and make conversations more natural than with the usual voice.
Alexa mimics voice in less than a minute
With Alexa’s new voice imitation feature, Amazon is taking another step in exploring artificial intelligence. Presented in a video from the re:Mars conference keynote, this technology demonstrated its ability to analyze and quickly copy voices.
Artificial Intelligence Act: why should non-discrimination be at the center of the project?
During the presentation, Amazon gave a controversial demonstration where a child asked the assistant to read The Wizard of Oz in the voice of his late grandmother. The voice assistant applied itself to this by appropriating the voice of the little girl’s grandmother to tell the story.
” This required inventions where we had to learn how to produce a high quality voice with less than a minute of recording against hours of recording in the studio said Rohit Prasad, Senior Vice President and Chief Scientist of Alexa.
Keep close with voice
Amazon wants to keep people close to their families, friends and relatives, with the voice of Alexa. Thanks to this function under development, they will be able to have a natural conversation.
The American company believes that Alexa offers a human dimension, which is essential in a difficult period, such as the health crisis. Despite separations, faithfully reproducing the voices of loved ones can bring relief and support, says Rohit Prasad. ” If AI can’t take away the pain of loss, it can arguably make their memories last “, he continued.