The American company Meta has introduced MusicGen, a new generative artificial intelligence that can generate music based on text input. The AI was created by Audiocraft's internal staff and can compose 12-second instrumental pieces.
The AI was trained on 20,000 hours of music from more than 10,000 licensed music tracks, as well as 390,000 free instrumental compositions from the Pond5 and Shutterstock media libraries, according to the creators. To get a better outcome, upload a song which reflects what you desire to achieve. MusicGen is available for free on the Hugging Face website. Google previously introduced MusicLM, a neural network that generates music from a text description, which is now available via the AI Test Kitchen app for mobile devices.
MusicLM also has a number of ethical difficulties, including copyright violation, according to Google researchers. During the trial, they discovered that approximately 1% of the music created by the system was merely extracts from the songs on which it was trained.
Photo: Kay Ibrahim
The Music industry has mixed feelings about AI entering their field. Alfa Mist, a jazz musician and producer, told Sky News that he didn't think it will have an impact on his music since his style requires "human error".
"I need a human being's experiences because that plays into the music," he adds, stating that he'd only consider utilising AI to "aid my creative process" in areas like mixing and mastering. “When tools get made, they can be of use or they can be exploited... regulate it before we're all finished,” adds Alfa.
Sting, however, in his interview in May of 2023, "The building blocks of music belong to us, to human beings," he explained to the BBC.
"Defending our human capital against AI will be a battle we all have to fight in the coming years," says an English musician and actor.
Comments (0)