Development of predictive model of demo singers' work efficiency using adapted perceptual evaluation of singing quality framework in the music recording industry in China

Demo singers are vocalists who record song demonstrations in recording studios before the songs are officially sung or dubbed by contract artists for release. This category also includes singers who perform for teaching materials. The, requirement of the efficient voice selection and information gap between music education and industry, along with the integration of Artificial Intelligence (AI), have prompted an exploration into the role of “hidden” singers in the music industry. Methods: This study adopts a mixed method approach, involving three sessions: observation of recordings, extraction of vocal parameters, and interviews with stakeholders. The research includes ten experienced singers (N=10) with a minimum of eight years of experience, who sang the same original song at Weake Studio, China. The vocal recordings were divided into pieces (N=380), and vocal parameters were extracted using the PESnQ framework. Additionally, interviews were conducted with sixteen stakeholders from four different production teams, each providing unique perspectives on demo singers. Results: The findings revealed outstanding pitch control abilities and personalized dynamic ranges among demo singers. A predictive model was proposed to improve the accuracy of artist efficiency prediction, further exploring parameter directions for future music production and AI voice humanization. Qualitative data analysis identified three levels of output from current musicians: technical output, emotional output, and cultural output. Implications: These discussions offer valuable insights to enhance the quality of music content production and promote sustainability in the music industry. Moreover, the study proposes recommendations for potential pedagogical research that align with the current state of the industry.

Text 122755.pdf Download (1MB) Official URL or Download Paper: https://ethesis.upm.edu.my/id/eprint/18657

Actions (login required)

View Item