Singing and listening to music are fundamental human activities that reflect society and culture. The emergence of vocal synthesis technology has expanded the landscape of music creation, with Vocaloid—exemplified by Hatsune Miku—gaining widespread popularity in Japan. While human singers and Vocaloids perform songs, their lyrics differ in thematic content, reflecting distinct creative and cultural dynamics. This study analyzes the lyrical trends of Vocaloid and human-performed songs from 2008 to 2023, focusing on topic distribution and temporal changes. Using BERTopic for topic modeling and a large language model (LLM) for topic labeling, we examined 800 Vocaloid songs from Niconico Video and 796 human-performed songs from the Billboard Japan Year-End charts. The results reveal that Vocaloid lyrics frequently contain negative themes, such as “lie” and “fool,” while human lyrics emphasize romantic and optimistic themes. Additionally, human lyrics exhibit greater fluctuations over time, particularly in response to social events such as the COVID-19 pandemic. Our findings highlight the distinct evolution of lyrics between Vocaloid and human singers, emphasizing the role of computational technologies in shaping modern music culture. This research contributes to understanding how vocal synthesis influences lyrical expression, expanding the boundaries of creativity and cultural transformation in music.