2024
Reducing Barriers for Marginalized Music in AI
This 4-month project explored barriers to the use of marginalized musical genres in AI music making in the UK and China. The Creative Industries and Intangible Cultural Heritages of the UK and China provide a rich ecosystem of creative industries AI research, along with substantial cultural heritage beyond the dominant forms typically used in current music AI research. Current deep learning approaches to music generation typically rely on extremely large musical datasets to train the AI models, which creates barriers to generating music beyond the dominant musical genre datasets, such as Western Classical music or pop music. For example, it is not possible to train these models on many minority culture genres, such as Qin genre in China, nor contemporary subcultures such as glitch or algorithmic music.
As part of the project team, I created a presentation video that demonstrates how to train an AI model (RAVE 2.3) using marginalized music datasets and how to use the model for composition. Additionally, I composed the music featured in our project’s documentation, blending traditional elements with innovative AI techniques.
As part of the project team, I created a presentation video that demonstrates how to train an AI model (RAVE 2.3) using marginalized music datasets and how to use the model for composition. Additionally, I composed the music featured in our project’s documentation, blending traditional elements with innovative AI techniques.
AI Music Composition
We commissioned a new piece of music generated using an AI model (RAVE 2.3) trained on a dataset of traditional Chinese musical instruments from the Chinese Musical Instrument Database (中国乐器数据库). Using 36 hours of audio from the dataset 108 hours of RAVE model training was undertaken with an A100 GPU.Video guide to using RAVE 2.3 with a dataset of traditional Chinese musical instruments from the Chinese Musical Instrument Database (中国乐器数据库).
Lead
Prof. Nick Bryan-Kinns | University of the Arts London, UK
Prof. Zijin Li | Central Conservatory of Music, CN
Music and Video
Zhou Zhou
Funded by AHRC-Funded Fellowship to Explore UK-China Creative Industries Research and Innovation Hub.