Developers listened to millions of audio clips to train the Fugatto model.
Researchers have developed artificial intelligence that they claim has created sounds never heard before.
This is reported by The Independent.
The AI model, called Fugatto, was developed by a team at computer chip maker Nvidia and aimed to create a “Swiss Army knife for audio.” Fugatto is used to edit or generate audio using text prompts.
Some of the text prompts, for example, remove certain instruments from songs or change someone's accent. Nvidia's Richard Kerris said Fugatto is more agile than other AI models and could revolutionize music by creating new sounds.
He commented, “Fugatto can make a trumpet bark or a saxophone meow. Anything a user can describe, the model can create.”
In particular, Fugatto developed audio samples from the text prompt: “Deep, rumbling bass pulses paired with intermittent, high-pitched digital chirps, like the sound of a massive intelligent machine waking up.” In another example, AI transformed the sound of a train into a string orchestra.
Ido Zmishlani, a producer involved in the Nvidia Inception program, added: “This thing is wild. Sound – it's my inspiration. It's what drives me to make music. The idea that I can create completely new sounds on the fly in the studio – it's incredible.”
Fugatto took developers more than a year to create, and millions of audio samples were used to train it. While AI has demonstrated its ability to create incredible sounds and videos with minimal effort, there are concerns about how the technology could impact creators and their work.