Nvidia Unveils AI Model Crafting Unprecedented Sounds
Nvidia has unveiled its new AI audio model, Fugatto, which can synthesize entirely new sounds by combining various audio traits and descriptions. Fugatto uses advanced training methods and a system called ComposableART to create unique audio outputs, such as a saxophone barking or a violin sounding like a laughing baby. The model is built on a heavily annotated dataset of 20 million samples and 2.5 billion parameters, allowing it to perform complex audio tasks and offer customizable soundscapes. Nvidia envisions Fugatto as a tool for audio artists, enhancing creativity in music and sound design.