MusiciAI: A Hybrid Generative Model for Music Therapy using Cross-Modal Transformer and Variational Autoencoder

Music has long been recognized for its therapeutic potential, offering emotional and psychological benefits. In the contemporary landscape, the intersection of artificial intelligence and music therapy introduces innovative applications. MusiciAI introduces an advanced generative model designed to harness the healing power of music through a novel hybrid approach. Combining a Cross-Modal Transformer and a Variational Autoencoder, MusiciAI interprets textual prompts and generates diverse, emotionally resonant music. The proposed model is trained on a carefully curated dataset, incorporating language models for precise annotation. MusiciAI integrates user interaction, allowing prompt-based music generation and user adjustments for a personalized therapeutic experience. Through a comprehensive exploration of the hybrid model's architecture, training, evaluation, and user-centric features, and contributes to the evolving landscape of generative AI in music therapy.