[post_excerpt]
A random noise signal, music and/or an encoded text prompt
A structured and high-fidelity music composition
To generate high-quality, coherent, and stylistically diverse music that aligns with user preferences or given constraints.
Trained on large-scale datasets of music, employing denoising and stepwise refinement techniques to enhance audio generation quality.
Measures include perceptual audio quality, coherence with input prompts, diversity of generated compositions, and user preference scores.
EN 
ES