A new project uses artificial intelligence to generate songs in the styles of deceased musicians, raising awareness about mental health in the music industry. The project features AI-generated music in the styles of Jimi Hendrix, Amy Winehouse, Jim Morrison, and Kurt Cobain, among others.
AI-Generated Music: A Tribute to Fallen Musicians
Since Kurt Cobain’s death in 1994, fans have speculated about his potential future work. Beyond the final Nirvana recordings and a few comments about collaborations, much remains unknown. To mark the anniversary, an organization created a new Nirvana song using AI, aiming to capture Cobain’s distinctive style.
Nirvana: Drowned in the Sun
The AI-generated Nirvana song, “Drowned in the Sun,” incorporates various guitar riffs, ranging from the calm style of “Come as You Are” to the raw intensity of “Scoff.” The lyrics, such as “The sun shines on you but I don’t know how,” reflect Cobain’s evocative style. While the vocals are performed by Eric Hogan of a Nirvana tribute band, the song’s creators claim that the music and most lyrics are computer-generated.
This initiative aims to highlight the tragedy of Cobain’s suicide and encourage musicians to seek help for mental health issues like depression and anxiety. “Drowned in the Sun” is part of a larger project, “Lost Tapes of the 27 Club,” which also features AI-generated songs in the styles of other musicians who died at age 27.
The project is a collaboration with Over the Bridge, a Toronto-based non-profit supporting music industry professionals struggling with mental illness. Sean O’Connor, Over the Bridge board member and creative director at Rethink advertising agency, posed the question: “What if all the musicians we adore had access to mental health support?” He highlighted the often-accepted and even glorified nature of depression in the music industry, suggesting that music sometimes represents genuine suffering.
The AI Songwriting Process
To create the songs, O’Connor and his team utilized Google’s Magenta AI tool. Magenta learns to compose in the style of specific artists by analyzing their existing work. The program has been previously used by Sony for a “new” Beatles song and by the electropop group Yacht.
Magenta analyzed artists’ songs as MIDI files, converting pitch and rhythm into digital code. The computer analyzed note choices, rhythmic patterns, and harmonic preferences, then generated new music. The team then selected the most promising musical sections.
O’Connor explained that more MIDI files improve the results. They used 20 to 30 songs per artist, focusing on hooks, solos, vocal melodies, and rhythm guitar parts. Processing entire songs proved less effective than using collections of riffs, which produced around five minutes of new material, most of which was unusable.
An artificial neural network AI was used for lyrics. By inputting existing lyrics and a few starting words, the program predicted the poetic cadence and tone. The team reviewed many pages of generated lyrics to find suitable phrases that matched the AI-generated vocal melodies.
Amy Winehouse: Man, I Know
Once the AI-generated musical sections were complete, an audio house arranged them. O’Connor noted that much of the instrumentation was MIDI with various effects added. Tribute artists were chosen as vocalists due to their ability to emulate the original artists’ inflections.
The Role of Human Musicians
Eric Hogan, frontman for the Atlanta-based Nirvana tribute band Nevermind: The Ultimate Tribute to Nirvana, provided vocals for “Drowned in the Sun.” Initially skeptical, Hogan was convinced by the quality of the AI-generated music. He found the AI-generated music challenging to interpret, requiring guidance from the project’s creators to understand the intended performance.
The entire project, from research and development to recording, took about 18 months. The team consulted superfans to identify potential plagiarism, particularly for the Doors-inspired song, “The Roads Are Alive,” ensuring it didn’t unduly resemble “Peace Frog.”
The Doors: The Roads Are Alive
Nirvana proved challenging for the AI, due to Cobain’s use of chord progressions that confused the algorithms. O’Connor described the resulting AI-generated Nirvana music as a “wall of sound.” Hogan, however, felt the final product captured the essence of Nirvana while avoiding legal issues.
Hogan praised the AI-generated lyrics for their directness while maintaining Cobain’s characteristic style. He believed the lyrics conveyed a sense of self-acceptance despite acknowledging personal struggles. He saw it as a statement of “I’m a weirdo, but I’m okay with it.”
Hogan also commented on the ethical implications, acknowledging uncertainties surrounding the use of AI in music creation. He pointed out the paradoxical nature of his role as a tribute artist.
Jimi Hendrix: You’re Gonna Kill Me
O’Connor anticipated criticism from those who view AI music as a threat to human artists. However, he emphasized the project’s positive intent and the significant human involvement in the process. He was confident in the quality of the result, suggesting the project successfully blended human creativity and AI technology.
Promoting Mental Health Awareness
Over the Bridge uses various methods to raise awareness and provide support for mental health, including Facebook groups, Zoom sessions, and workshops to combat isolation among artists. The project has no current plans to sell the generated tracks. Michael Scriven of Lemmon Entertainment highlighted the value of simply acknowledging shared struggles.
Scriven also emphasized the extensive human effort involved in AI music creation. He pointed out that the notion of AI replacing artists is currently misleading, requiring considerable human input throughout the entire creative process. Each song involved several collaborators, dispelling the idea of simple button-push music creation.
Read also: AI DJs: A Threat or Tool? Understanding Copyrights & Fair Use
* generate randomized username
- COMMENT_FIRST
- #1 Lord_Nikon [12]
- #2 Void_Reaper [10]
- #3 Cereal_Killer [10]
- #4 Dark_Pulse [9]
- #5 Void_Strike [8]
- #6 Phantom_Phreak [7]
- #7 Data_Drifter [7]
- #8 Cipher_Blade [6]