New Delhi: Researchers at US tech giant Google have created an AI that can generate minute-long musical pieces from text prompts, and also convert a whistle or hum to other instruments, as written by systems such as DAL-E. The signals generate images, reported The Verge, an American technology news website, via TechCrunch.
According to the outlet, the model is called the Music LM, and while you can’t play with it for yourself, the company uploaded a bunch of samples it created using the model. The examples are impressive. ,Also read: LIC Policy: Invest Rs 1300 every month, get Rs 27.60 lakh, get it like this,
There are 30-second snippets of sound like actual songs built from paragraph-long descriptions of a genre, vibe and even specific instruments, as well as five minutes generated from one or two words like “melodic techno”. Let’s also determine the long pieces of . ,ALSO READ: Apple’s Emergency SOS via satellite rescues two women stranded in Canada,
Also featured on the demo site are examples of when 10-second clips of instruments like cellos or maracas, eight-second clips of a certain genre, music that fits a prison escape, and even what Makes a model sound like an advanced to novice piano player.
It also includes interpretations of phrases such as “futuristic club” and “accordion death metal”, reported The Verge. MusicLM can also emulate human vocals, and while it seems like the tone and overall sound of voices are accurate, there’s a quality to them that’s definitely off.
According to The Verge, AI-generated music has a decades-long history; There are systems that have been credited with composing pop songs, imitating Bach better than humans, and performing live in the 90s.