With the proliferation of AI music generation tools (e.g., Suno, Udio), non-expert users face a significant gap between emotional expression and musical parameters: they know what emotion they want to convey but lack the knowledge to translate it into music terminology. This paper proposes a systematic Emotion-Oriented Prompt Library that maps users' natural language emotional expressions to structured musical parameters, covering instrument selection (60+ types), genre styles (80+ types), narrative moods (40+ types), and vocal styles (30+ types). By providing "emotional tags" and "one-sentence personality descriptions" for each musical element, this library significantly reduces the cognitive barrier for ordinary users. Preliminary user feedback indicates that the success rate of generating emotion-aligned music increased from approximately 40% to over 85% after adopting this library.
xiaozhou zhou (Sat,) studied this question.