This piece was premiered at the SCOPE III concert in Wu Performance Hall at UC Berkeley on November 15, 2024. The fragments of video in the piece are all generated from a custom software I created that uses Stable Diffusion to generate music reactive videos based on text prompts with time stamps. Each video is the result of the same series of prompts, but with different random seeds, giving each a slight variation. At certain points, a laser begins to draw on the projection screen, overlaying patterns and shapes onto the generated videos.
The smallest things, like grains of sand and cells, and the largest structures, like the cosmos, the “celestial machine,” are all created from the same fundamental building blocks. The title of the work, drawn from Leonardo da Vinci’s notebooks on water, points to a tension between the concepts of breathing and machine. Breathing is not just a fundamental aspect of being alive; it offers us a path to turn inward and even a connection to our spiritual life. How can we reconcile breathing, something so close to our humanity, with the seeming artificiality of machines? If this "celestial machine" is truly a machine at a grand scale, what does its breathing look like?
Technical details on the video generation software can be found in my paper "Network Bending of Diffusion Models for Audio-Visual Generation" in the 2024 proceedings of the Digital Audio Effects Conference (DAFx).
Recordings of modular synth improvisations are input to an oscilloscope which visualizes the waveform in real time.
Premiered April 17, 2023 by Richard Worn at the Center for New Music and Audio Technologies (CNMAT) at UC Berkeley. Created from collaborations with bassist Richard Worn, From within explores the frequency range of the bass by pairing the rich low frequencies that the instrument is known for with the high harmonic spectra that result from different bowing techniques. The electronic accompaniment is made entirely from recordings of Worn playing the bass, producing a personal sonic environment that envelops the instrument and performer.
Made using only two sine waves and two sawtooth waves, electronic sounds are visualized in real time through an analog oscilloscope. By changing the frequency relationships between the left and right audio channels, different Lissajous patterns emerge.
A large, clay statue of the Buddha sat in a temple in Thailand’s Sukhothai province for hundreds of years. The statue was revered for its long life, but as time went on the hot weather caused the clay to crack and break. One day, a monk peered into one of the largest cracks, and from deep within the statue shined the glint of gold. The statue had been covered with layers of clay to protect it during times of unrest, but the knowledge of its true nature had been forgotten for over two hundred years
This piece was not written by me. I did not write the text. I did not write the music. I did not set the text to the music. All of these were done by various computer programs.
The lyrics, which were written by a neural network named GPT, are:
This is a song, written in the year 1537. It was written by a guy named Philip Neri, in the city of Naples, in Italy. It was a religious piece. He was a priest. He was a composer, a guy who liked to make music and sing. A lot of the music of that time was quite complex and had a lot of rhythm and structure. But it was the most common way of communicating ideas back then. There were no instruments. Just a lot of singing. This is the piece I created using these lyrics and that sounds a bit like some kind of early church hymn. The chords are based on the Ionian scale. The tempo was around 130 BPM and it takes about two minutes to play this piece. This is
This work is an exploration of questions of machine creativity. Can computers be creative? How do we value art differently based on whether it was made by a human or a machine? What is happening when a program creates a piece of music? Ultimately, we have a historical re-reading of chant music through the lens of modern technology: neural networks, Markov chains, artificial intelligence. It’s what happens when you ask a computer to write a piece of music.