Home Music and Digital Media Ars Technica Nvidia’s new AI audio model can synthesize sounds that have never existed