YuE2

YuE2 unifies symbolic and audio music generation in one model, achieving frontier song quality competitive with leading platforms. Through symbolic planning, it writes an editable score, then brings the composition to life with vocals and accompaniment. The project also features advanced models for music representations (MERT2) and audio-to-score transcription (SheetSage2), offering comprehensive tools for music creation, analysis, and research. YuE2 is an open-source initiative, providing benchmarks and models on Hugging Face for further exploration and development.