Music recommendation systems play a crucial role in helping users discover new music that matches their references and vibe. This paper introduces a formula-based recommendation model designed to overcome the limitations of current music recommendation systems. A two-phased approach has been developed for new users where initial recommendations are generated based on a popularity score derived from aggregated listening trends and cosine similarity algorithms, followed by lyrical analysis using natural language processing to suggest thematically relevant tracks. For existing users, the model leverages historical interaction data to compute a similarity recommendation score, incorporating personalized user profiles, real-time ratings, and a novelty factor to suggest new and interesting tracks. Unique parameters such as listening time and repeat rate of a song play a crucial role in enhancing the recommendation accuracy. Listening time captures the user's engagement level with a particular song, providing insights into their preferences beyond just skips and likes, whereas the repeat rate of a song indicates its popularity and relevance, reflecting how frequently a user revisits it. By combining these elements, the system aims to improve music discovery, engagement, and user satisfaction. This paper discusses the design, implementation and recommendation system, evaluation demonstrating effectiveness in delivering personalized music recommendations to users.