Music generation is a challenging task that requires capturing the complex and diverse aspects of musical structure and expression. In this paper, we investigate the factors that affect the quality of music generated by various AI models, such as MuseGAN, MuseGAN-Image and GPT3-Music¹[1]. We use different data encoding and processing techniques to create and evaluate music generation models based on generative adversarial networks (GANs) and transformers. We compare the advantages and disadvantages of each method in terms of harmonic, temporal and spatial aspects of music. We identify several challenges and drawbacks of the existing methods, such as harmonic loss, GAN overshooting, chord progression, octave representation, and framework compatibility. We also suggest some possible solutions and future directions for improving music generation with AI.