NVIDIA Launches Fugatto: The 2.5B-Parameter Audio AI Revolution

In late 2024, NVIDIA unveiled Fugatto, a groundbreaking generative AI with 2.5 billion parameters, trained on a DGX system featuring 32 H100 GPUs. Fugatto can craft and transform music, voices, and sounds from text or audio inputs—imagine a trumpet that meows or vocals spun from piano notes. Aimed at music producers, advertisers, and game developers, NVIDIA is delaying public release to address ethical and copyright issues. Fugatto marks a significant leap in multimedia AI, promising to reshape creative industries with a dash of playful sound innovation.