Prompt-based music generative artificial intelligence (GenAI) offers an efficient way to engage in music creation through language. However, it faces limitations in conveying artistic intent with language alone, highlighting the need for more research on AI-creator interactions. This study evaluates three different interaction modes (prompt-based, preset-based, and motif-based) of commercialized music AI toots with 17 participants of varying musical expertise to examine how prompt-based GenAI can improve creative intention. Our findings revealed that user groups preferred prompt-based music GenAI for distinct purposes: experts used it to validate musical concepts, novices to generate reference samples, and nonprofessionals to transform abstract ideas into musical compositions. We identified its potential for enhancing compositional efficiency and creativity through intuitive interaction, while also noting limitations in handling temporal and musical nuances solely through prompts. Based on these insights, we present design guidelines to ensure users can effectively engage in the creative process, considering their musical expertise.
https://dl.acm.org/doi/10.1145/3706598.3713762
The ACM CHI Conference on Human Factors in Computing Systems (https://chi2025.acm.org/)