Prompting for Voices (2024)

Prompting for Voices is an experimental project that examines how commercial AI models interpret lesser-known ethnic music. The goal was to generate traditional folk music from around the world using AI models and analyze how well these models, primarily developed with Western-centric datasets, represent minority cultures.

The final output is a non-narrative documentary-style video artwork, where AI-generated content is juxtaposed with original recordings and visuals. The process involved frame interpolation and inpainting techniques to seamlessly blend AI-generated sections with original footage and audio. Models such as Runway and Suno were used to generate video clips and sound through prompting.

This project explores the music and visual culture of diverse ethnic groups, including the Cree, Miao of China, Yoruba, Bunun of Taiwan, Korean Ganggangsullae, and Mongolian music. At the end of the video, the actual recordings of these ethnic groups are presented, allowing viewers to compare the AI-generated versions with authentic representations.

Group Project

My role: Content Researcher, Visual Creator

Team: 2 Audio Creators