Nvidia’s new AI music editor, Fugatto, can create unique sounds, music and speech from text and audio prompts it hasn’t been trained on. The tool generates imaginative outputs, like a “saxophone howling, barking, then electronic music with dogs barking,” or transforming voices by altering accents or tones. It can also edit music, isolating vocals, adding instruments or swapping melodies.
Fugatto was trained on a vast dataset, including BBC sound effects, and uses advanced instructions to expand its capabilities without extra data. While similar tools exist, Nvidia claims Fugatto creates completely novel sounds. Availability details remain unannounced.
Source: The Verge