YuE

YuE2 is a cutting-edge music generation system that unifies symbolic and audio creation, delivering frontier-quality songs with editable compositions. Users input lyrics and a style, and YuE2 crafts an editable melody and chord plan, then renders it into a complete song with vocals and accompaniment. It features white-box generation, allowing inspection and modification of the symbolic plan before audio synthesis. YuE2 also supports zero-shot covers—reimagining existing songs in new styles—and agentic editing, where users refine compositions through interactive conversations. Its architecture employs an AR–NAR Mixture-of-Transformers backbone for score and semantic token prediction, generating acoustic latents via flow matching, which a VAE decodes into stereo audio. It rivals leading models like Suno v5/v6 in quality.

• Frontier-quality unified symbolic and audio music generation
• White-box editable melody and chord planning
• Zero-shot song covering and agentic interactive editing
• Based on AR–NAR Mixture-of-Transformers architecture
• Integrates as an OpenClaw agent skill