NVIDIA Introduces Fugatto AI Model with 2.5B Parameters for Audio Generation and Modification

Nov 25, 2024, 05:47 PM

NVIDIA has unveiled Fugatto (Foundational Generative Audio Transformer Opus 1), a new generative AI model with 2.5 billion parameters designed to generate and modify audio—including music, voices, and sounds—from text and audio prompts. Announced on November 25, 2024, Fugatto can create any combination of music, voice, and sound, transforming sounds in innovative ways such as making a trumpet bark or a saxophone meow. The model, not yet publicly released, was trained on open-source data and is aimed at professionals in the music, film, and gaming industries. According to NVIDIA's Richard Kerris, Fugatto can modify voices, change accents, adjust emotional tones in speech, add instruments to existing music, and generate unique soundscapes, representing NVIDIA's push into audio creativity and innovation.