YouTube has sidestepped controversies surrounding the use of artificial intelligence (AI) to generate new music with a new tool that clones singers’ voices.
The new feature, called Dream Track, is available in YouTube Shorts – the Google-owned platform’s answer to TikTok that lets users post short videos.
Users can enter a prompt about what sort of music style they want – such as ‘upbeat’ or ‘ballad’ – and select the artist they want the AI to imitate.
Nine artists have allowed their voice to be copied for the tool, including Alec Benjamin, Charlie Puth, Charli XCX, John Legend, Sia and Troye Sivan.
YouTube posted a short clip of what the clone version of US singer Charlie Puth sounds like – and it’s impressively close to the real thing.
Dream Track is available in YouTube Shorts – the platform’s answer to TikTok that lets users post videos up to 60 seconds long
The AI song generator sings a fresh set of lyrics too: ‘Baby, we’ve got nothing in common, but I know I’m what you’ve been wanting.’
Dream Track is the creation of DeepMind, the London AI company founded in 2010 and bought out by Google in 2014 for $500 million.
In a blog post, YouTube staff said AI has ‘brought music to the cusp of a new creative era’ and that it can ’empower creativity’.
‘When technological innovation, human imagination and music meet, extraordinary things can happen,’ they said.
‘It’s still early days but we are motivated by our progress in AI and music and are energized by the possibilities that lie ahead.’
Dream Track is currently only being tested with ‘a small group of select US creators’ and is not available to the general public yet.
MailOnline has contacted Google about when users in the UK will be able to try it out.
Users need to type in a prompt requesting the style of music and lyrical content before selecting an artist to hear the song in their ‘cloned’ voice
Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Papoose, Sia, T-Pain, and Troye Sivan who have chosen to collaborate in this experiment
With Dream Track, users can create soundtracks of up to 30 seconds for their YouTube Shorts.
They first need to type an idea into the creation prompt – such as ‘a ballad about how opposites attract’ that is ‘upbeat’ and ‘acoustic’.
They then need to select one of the nine artists who have agreed to collaborate in this ‘experiment’ from a carousel.
‘An original Shorts soundtrack featuring the AI-generated voice of that artist will be produced for the creator to use in their Short,’ YouTube said.
It’s unclear how much Puth and the other artists have been paid for the use of their likeness, but the New Jersey singer appeared excited about the tech’s possibilities.
He said: ‘YouTube has been a great partner in outlining their approach to AI and understands the need to work together to develop this technology responsibly, ensuring it will accelerate creativity instead of replacing it.
Meanwhile, Florida rapper T-Pain said he’d ‘always been about pushing the boundaries of technology and harnessing it to make the most interesting music for my fans’.
Florida rapper T-Pain (pictured) said he’d ‘always been about pushing the boundaries of technology and harnessing it to make the most interesting music for my fans’
AI tools such as Dream Track and OpenAI’s Jukebox are fed an artist’s songs to ‘learn’ their vocal styles and musical hallmarks and be able to generate brand new approximations, with new lyrics and melody.
Some musicians are excited about the prospect of being able to offload at least part of the songwriting process to a machine.
Earlier this year, British pop icons the Pet Shop Boys argued that AI can be used in a positive way in the creative process.
The group’s singer, Neil Tennant, said AI could ‘fill in the blanks’ if a song has been left unfinished, such as when the composer is suffering from writer’s block.
However, others are far more wary as the technology can throw up problems with music copyright and undermine artistic integrity.
AI-generated music blurs the line between using a song protected by copyright and using a cheaper or copyright-free approximation.
Neil Tennant, performing here in Milan in May 2022, says AI could be a tool for songwriters to create music
Recently Canadian rapper Drake has been victim of AI songs developed to sound like him, which his record company have been trying to put a stop to.
‘If someone hasn’t used the actual recording you’d have no legal action against them in terms of copyright with regards to the sound recording,’ said Rupert Skellett, head of legal for British record company Beggars Group.
However, not all forms of AI in the music industry have been built to digitally replicate the sound of artists in this way.
For example, Paul McCartney used an AI tool called MAL to isolate John Lennon‘s voice from a 1970s home demo and render it ‘crystal clear’.
MAL – which can be trained to identify different instruments such as guitar, piano and vocals and separate them – let him and Ringo Starr finish the ‘last Beatles song’, which hit the top of the singles chart last week.
