Nvidia launches Fugatto AI model to create unique soundscapes

Nvidia has launched a new AI model called Fugatto, which creates unique soundscapes by transforming music, voices, and other sounds. This model uses advanced training methods and has 2.5 billion parameters, trained on over 50,000 hours of audio data. Fugatto employs a technique called Composable ART, allowing it to generate novel sound combinations based on text or audio prompts. For instance, it can produce a violin sound resembling a laughing child or a factory machine that screams. In addition to creating new sounds, Fugatto can modify existing audio by changing emotions, isolating vocals, and adapting instruments. This marks Nvidia's shift from solely providing hardware to developing innovative AI technology.