This AI Songwriting Platform Can Imitate Big Pop Stars

For those who can't wait for the next big release from their favorite artist, British website builder Zyro is attempting to fill that gap with artificial intelligence. The company launched a platform this week that automatically generates sometimes-passably-realistic lyrics in the style of Taylor Swift, Kendrick Lamar and The Beatles as a way to promote its machine learning-powered content generation tools. Built on a groundbreaking text generator from research group OpenAI called GPT-2, the algorithm is trained on 50 hours of music and lyrics from each of the above artists. Zyro even brought to life one of the generator's catchier outputs, a song from AI Kendrick Lamar titled "Revivals" that includes these lyrics: Dark Messiah, you're Richard Pryor's Motown gang (Reverber!) I need my break to go back to the real quick (Yeah) We'll be rockin' like revivals (Oh), Re-we'll be rocking like revivals (Reverber!) [kickout id="1"] Or, for a rawer example, our random sampling of the tool produced these Taylor Swift lyrics: You said she's like a shooting star 'Cause, she must be a bullet dream, what are you waiting for? Like a shooting star Like a shooting star You gave me the chance to make you mine You gave me the chance to take it too far, too far The songwriting platform is designed to promote Zyro's copy generation suite, also built on GPT-2, which can produce generic text content for different types of businesses ranging from bakeries to gyms and lifestyle digital media sites. It even generates random business name and slogan ideas. "This studio has a great location for an intimate workout," the tool wrote, when prompted to write copy for a gym, "along with a welcoming atmosphere for our community. Come explore our boutique studio with a variety of classes for all levels." The AI also suggested a laundry list of possible names for this hypothetical gym such as Questic and Xorp. [kickout id="2"] [article_summary section="1"]Tools like that make Zyro one of the first companies to attempt to commercialize recent breakthroughs in machine learning-powered text generation that have thus far mostly fueled novelty projects like parody Twitter accounts and a cult-hit strategy game (as well as Adweek's Shorty Award-winning Super Bowl Bot.) But it's not the only company trying to harness the often-chaotic power of these freeform text generators. OpenAI recently announced that its latest, much more powerful release, GPT-3, would serve as the backbone for its first commercial product, which has already lined up clients like Reddit and Quizlet. Amazon has also begun to test a chatbot based on similar technology that is able to generate wholly original dialogue. Zyro isn't the only company applying this technology to the realm of music either. OpenAI recently released a much more comprehensive AI music generator called Jukebox that can create fully realized tracks and lyrics in a variety of genres. Agency Space150 based a campaign around an AI version of Travis Scott last year, and visual editing platform PicsArt recently launched an AI music generation feature as a way to circumvent royalties. [kickout id="3"] “Several creative industries, including the music industry, are fearful of AI and believe it has the ability to replace their jobs," said Zyro's head of AI, Tomas Rasymas, in a statement. "But when it comes to the music industry, it’s more likely that we’ll see the technology being used to open up new opportunities for musicians rather than threatening to completely replace them. "Rather than taking over human songwriters, AI is set to expand the field for new artists, making music creation more accessible. It may even allow existing artists to experiment with AI and create alternative forms of music altogether," Rasymas added. [kickout id="4"]

AI News

This AI Songwriting Platform Can Imitate Kendrick Lamar or Taylor Swift

The algorithm was created by U.K. website builder Zyro

Patrick Kulp

About

Subscriptions

Events

Publications

WORK SMARTER - LEARN, GROW AND BE INSPIRED.

Subscribe today!

Patrick Kulp