[To hear one of Chen Yu-Han’s complete recordings of my solo work “Hook” please go to my bandcamp page here. More recordings, including fragments and improvisations, to follow!]
In the last year and half I’ve been looking into approaches to human-AI co-creation, working in particular with Glasgow-based scientist Lex O’Brien. They have been helping me determine a viable methodology, how much recorded material is required to “feed” to the AI, and potential research questions and creative outputs.
I have previously focussed on using my Yayue compositions as the basis for research, but we became aware of practical and technical issues that would hamper research in that area, namely the large number of performers potentially required, and the sheer timbral variety of all the instruments in a Yayue ensemble. While the first issue could potentially be solved by my playing the instruments myself, and recording them individually and mixing them to get the effect of a full ensemble, this would be very work intensive and would involve my learning to play all those instruments, some of which are harder than others. The second issue would complicate the process of training the AI to assimilate the sound-world and typical stylistic behaviours of the music, thus requiring many more hours of recorded material and many more hours of compute time.
Lex and I have settled on an alternative (interim?) project that would be more manageable as a starting-point on this vastly ambitious journey: instead of Yayue we shall use the music of the guzheng, and specifically my composition for solo guzheng, Hook. To this end we have teamed up with the very young and talented guheng player Chen Yu-Han, to record both Hook and a set of related materials as the basis for training the AI. With just one instrument- albeit one capable of a vast array of different performance techniques and sonic effects- this should afford us a greater probability of the AI generating intelligible (and interesting) results.
I need an hour of music for the training process to work, whereas Hook is just ten minutes’ duration. To expand those ten minutes of material to sixty minutes, I devised a number of strategies:
1. Get Yu-Han to play Hook using two additional scordatura, in effect generating two gesturally and rhythmically identical, but harmonically quite different pieces.
2. Record all the rehearsals with Yu-Han, thus allowing me to extract passages of us “trying-out” parts of the piece, which are also potentially valid as training material.
3. Getting Yu-Han to improvise on the seven different material types in Hook. Six of those materials represent my “transcriptions” of the playing styles of the six major schools of Guzheng playing, while the seventh is an “desolate and alienated” music played on the left-hand-side of the moveable bridges, which generates aleatorically microtonal musical material. I’m very excited about working with Yu-Han on how to approach these improvisations, and seeing what she comes up with.
The project has received a modest amount of funding from the Nanhua University Research & Development Department.
Guzheng related posts:
NEW WORK “HOOK” SELECTED FOR INCLUSION IN 2023 TAIPEI INTERNATIONAL NEW MUSIC FESTIVAL (Aug. ’22)
New work “Moon” selected for inclusion in 2022 Taipei International New Music Festival (Oct. 2021)
Opera for Children’s Choir: Cao Chong Weighs the Elephant 童声歌剧：曹冲称象 （2017）