ABOUT THE TEAM

Nous sommes Whim Therapy is a team composed of the artist Whim Therapy and several researchers from Sony CSL.

Whim Therapy is the solo project of Jeremy Bénichou, who is part of the successful French band Diva Faune. This project is an opportunity for Jeremy to explore new avenues, without caring too much about commercial success.

Sony CSL (Computer Science Laboratories) is a fundamental research laboratory based in Tokyo and Paris. In Paris, the “Music and Artificial Intelligence” team develops a new generation of music production tools based on Artificial Intelligence. When Whim Therapy heard about Sony CSL on Instagram, they connected and decided to enroll in this contest to seal the collaboration.

ABOUT THE SONG

This song is a soulful homage to old-school music, fulfilled with postmodern elements. The lyrics are all about addictions, expressed through love imagery. This song was conceived as a genuine collaboration between humans and machines. During early stages, we have paid attention to leave some room for the AI to play its part. It has done so wonderfully, leaving us with a song where analog and digital elements are blended to create a unique vibe.

ABOUT THE HUMAN-AI CO-CREATION PROCESS

We started the composition process the classic way: First, we chose a chord progression, a vibe, and a tempo. A first studio session allowed us to record guitar, piano, Rhodes, and toplines. Then we wrote the first three sentences of the song. Up to this point, AI was not involved. This first phase allowed us to create content that would be used as input to our various conditional models later on. We also wanted the foundation of this track to be as “empty” as possible to leave some room for the AI to feature in.

In a second phase, we started to involve the AI tools. We used BassNet to create a bassline that would fit the pre-existing content recorded during the first phase. We used DrumGAN to create the drum sounds. DrumNet helped in writing hi-hat and kick patterns. NOTONO was used to create the different synth sounds. Once we got all these, some heavy editing and processing was done to keep only the best elements.

The third phase consisted of a second studio session, where we re-wrote the toplines according to the new arrangement. We also used Poiesis to write the definitive lyrics, recorded final vocals, and run the stems through tapes to get this unique old-school flavor. Finally, heavy processing was added to the vocals, and voilà.