ABOUT THE TEAM

A/S (Artificial Spirit) is a team of composers, developers, and performers interested in exploring the vast world where art meets technology. Under the theme of ‘AI Meets Art’, we are working on a project to express experimental performance and art using machine Learning/Deep Learning and human emotion. Our attempts to do this can be found in one of our songs “Wheel Of Four Tune” to add emotion to music composed using Artificial Intelligence.

Our team’s goal is to inspire people by creating a genre of content where arts and technology work in harmony. We also aim to expand the art and technology convergence market through challenging trials and diverse analysis.

ABOUT THE SONG

Beyond the Kosmos is a song based on a concept derived from the Greek word kosmos (cosmos), which is a concept larger than the universe. The message that we want to send is that chaos should not shine beyond the universe as there are conflicts, disconnections and pain permeating humankind, and instead receive healing through ambient sound, which is heard in nature and daily life around us. This was made into a song with the hope that everyone would communicate, harmonize, and heal their wounds even in chaos. We wish everyone could feel the vast universe within their heart after listening to this song.

ABOUT THE HUMAN-AI CO-CREATION PROCESS

We divided the music generation into several stages and selected the part that can cooperate well with the AI. The process we chose was sound design, melody composition, and chorus.

First, we decided to create our own artificial intelligence instrument by collecting various sound data such as natural sound and ambient sound in order to express the concept of the song, the inner human and kosmos. In the case of the tone transfer technology, model training was possible with a data set of at least 20 minutes, so it was a suitable technique for conducting various experiments. A customized model was created using various sound data such as dolphins, bells, insects, birds, and water droplets, and tone transfer technology, and sound design was made by adjusting parameters. In the melody composition part, rather than random sound source generation, Magenta’s improv_rnn technology was used to generate MIDI files by receiving the chord progression intended by the composer as input. Lastly, when the vocal recording was performed using TTS technology, detailed vocalization and emotional expression were difficult, so we devised a strategy to use it as a chorus by bypassing it. As a result, it was possible to receive the help of artificial intelligence in various areas of music creation, such as the creation of a melody that maintains the context of the song, a new instrument that was a source of inspiration, and a number of choruses.