Music source separation is a critical task in audio processing and is essential for applications such as remixing, audio restoration, and music analysis. Traditional methods, including signal processing techniques like Independent Component Analysis (ICA) and Non-negative Matrix Factorization (NMF), as well as deep learning approaches utilizing Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), face challenges related to extensive data requirements, computational complexity, and limited adaptability to diverse audio contexts. Recent advancements in Large Language Models (LLMs) like GPT-4 introduce new possibilities for addressing these challenges without retraining models. This literature review explores the integration of prompt engineering and Retrieval-Augmented Generation (RAG) with LLMs to perform music source separation. By leveraging LLMs’ inherent language understanding capabilities and accessing external databases containing musical scores, lyrics, and instrument profiles, this approach aims to enhance efficiency, flexibility, and accessibility in source separation tasks. The review highlights how this innovative methodology can overcome the limitations of traditional techniques, filling a notable gap in current research and offering transformative potential for both academic inquiry and practical applications in music technology.