ABOUT THE TEAM

Zillion Inch Keytar is comprised of musicians and AI researchers working in the US Midwest with the intent of making engaging pop music infused at every level with AI technology. The band’s name—generated by an algorithmic band name generator—implies an almost fanatical approach to the use of AI. As such, parts of this bio have also been written with AI in stylistically aethsetic spaces within the context of palette-performer-tetrachords. This band, while primarily concerned with structured sound, also seek to re-decipher all ideas. In their most recent work, microtones, tones and coherent textures are all used, allowing the audience to seek a variety of simultaneous resonators. Audible choreography is the single most important element in any polyphonic composition, and this band’s work seeks to explore (and morph) this in the context of post-Webern consequences and many aleatoric hemiolas.

ABOUT THE SONG

This song was composed using AI, Machine Learning, and Algorithmic musical techniques to the fullest extent possible. The band’s name, song title, lyrics, melody, harmonic structure, drum track, images, mixing and mastering process were all performed using AI, ML or algorithmic software commonly available to the consumer. The title ‘Body of a Nightmare’ (suggested via an online song title generator) was fed into GPT-2, which can be used to do multitext training. In this case, team member Joel Matthys generated the lyrics from a dataset including the songs of Lennon & McCartney, Jewel, Judy Collins, and various advanced music theory textbooks. The lyrics determined much of the direction of the song.

ABOUT THE HUMAN-AI CO-CREATION PROCESS

The team initially decided upon a standard musical form, (Intro–Verse1–Chorus–Verse2–Chorus–Bridge–Chorus–Outro) and then generated material to fill out the various sections.  After the lyrics were generated using the GPT-2 multitext training process, the lyrics were inputted into a text-to-speech software program called Natural Reader 16, to assign a somewhat natural rhythm of the words (the voice of Natural Reader’s Moira text-to-speech synthesizer remains in the track toward the end on the line “You can be anything you want, there’s no need to think”). The audio from Natural Reader 16 was then transcribed using Ableton Live’s audio-to-MIDI melody algorithm specifically to notate the rhythm of each line. The team then used melobytes.com, a popular text-to-melody generator, to suggest melodic material. It was difficult to make this melody to fit with the predetermined rhythm of the text, so this stage saw the most human intervention. Once the melody was composed, the team used Magenta’s Transformer Collab notebook to generate chord accompaniments for each section. The vocal synthesis was completed using Synthesizer V’s Eleanor, the drum tracks were composed using Logic Pro Drummer, and the final draft of the track was mixed and mastered using iZotope Neutron Suite and Cubase.