A Comparative Evaluation on Melody Generation of Large Language Models

Music has always been central to human culture. With the rise of digital tech and smart devices, its role in the Consumer Electronics (CE) sector has grown. Recently, large-scale language models (LLMs) like ChatGPT and Bard are promising to change the way we compose music. This paper explores the composing ability of Large Language Models (LLM) such as ChatGPT and Bard. Unlike older methods that required deep musical and statistical knowledge, LLM lets users simply describe what they want in music. We evaluated this LLM on how well it matches user input and the overall quality of its compositions, looking at factors like repetitiveness and scale diversity. This research aims to highlight the strengths, uses, and limits of LLM in music, setting the stage for its greater role in CE and the wider music world.Generated Melodies: https://github.com/545659928/LLMMelody