Analysis of AI Singing Technology Based on Praat Software

This study explores the core technology behind many contemporary virtual singers, specifically AI singing technology, from the perspective of the Xiaoice framework. The paper compares AI singing with human singing in terms of resonance, breathing, head-chest voice transition, and articulation. Using Praat software, the study analyzes pitch, timbre, and transitions between vowels and consonants. The results reveal significant differences between AI and human singing in terms of pitch processing and articulation, which in turn affect the emotional expression and performance of songs. While certain features of AI singing are difficult for the human ear to distinguish, other features are easily identifiable. The study concludes that AI singing technology needs further refinement to surpass the expressive capabilities of the human voice. Additionally, the paper discusses the future prospects of AI singing technology, emphasizing that AI with singing capabilities, unlike AI that merely provides knowledge storage, is less likely to become obsolete and has considerable potential for development.