Music Source Separation (MSS) is a pivotal com- ponent of audio signal processing, committed to disentangling and separating individual sound sources from complicated audio combos. This paper provides an excellent method for music source separation by leveraging preprocessing strategies and data augmentation strategies such as time-stretching, pitch- shifting, background noise addition, and reverberation, our system enriches the training dataset for improved accuracy. The method employs Recurrent Neural Networks (RNNs) to decipher temporal dependencies and are to extract individual components from combined audio spectrograms. Guided by way of evaluation metrics such as signal-to-noise ratio (SNR), signal-to-interference ratio (SIR), this methodology achieves great precision. This paper’s findings signify improvements in audio sign processing, showcasing practical applications in numerous domains by disentangling complex audio combos to extract clearer and distinct sound sources