ABOUT THE TEAM

The word is that Transatlantic Waves was born out of a collaboration between Algoriffix and VoxGarten that was initiated by the computer scientist and music enthusiast Stanislaw Gorlow. After a successful first trial with the formally trained bass virtuoso and composer Juan Carlos Sardaneta and with the creative assistance of the music producer Jorge Costa, we decided to take on a bigger challenge by joining the AI Song Contest. To even out the team both geographically and musically, two more members from Stockholm came on board: Leo Krepper, an upcoming young guitarist with a strong background in jazz, and the multitalented musician and songwriter Adam Öhman, better known as Mountain Bird. Adam never fails to give a personal and human touch to electronic music. The team never worked together before, which seemed like adding some extra spice to the project.

ABOUT THE SONG

When we first sat down all together, virtually bridging the waves of the Atlantic with an ethernet cable, we wouldn’t have a clear idea where this journey would take us. Our first thought was to just go and try to create an AI version of a pop song with a funky bass line and a catchy hook. After a while, Adam came up with a strong demo and our vision crystallized. As we attach importance to the fact that human listeners can make a sense of the lyrics and possibly even relate or sing along, we refrained from AI gibberish. The song turned out to be about the imperfections of human nature, which become manifest in our emotions and endless desires and which seem hard to escape. Will our choices put us at the mercy of the artificial intellect...?

ABOUT THE HUMAN-AI CO-CREATION PROCESS

Given the musical potential of the team, it would have been a waste of frontal biomass to pin all our hopes on AI. Instead, we engaged in a human-centric workflow, which basically means that the creative process was driven by the team members while AI served as a “consultant” on all suitable occasions. Our aim was to explore how AI can help us create a pop song from scratch, not whether it could do all the fun stuff for us. For this, we mostly relied on Algoriffix’ tools for note recognition, musical analysis, auto-chords, and machine improvisation.

We started off with a simple four-bar guitar melody in common time and took it from there. AI figured out the key and the roots and recommended a chord progression. With a bit of chord juggling, we finally agreed on the harmony and the basic arrangement of the song. An early recording of the multilayered vocal inspired us to create an “AI solo” by converting the stem to MIDI. Since all AI data was in MIDI format, it was easy to develop the solo and other parts with standard DAW techniques such as articulation, quantization, or arpeggiation. The transcribed vocal was further used to assemble a body of short note successions from which a new melody that followed the bass line was improvised by AI. A mixing engineer gave the song a final polish.