PickYuE: Open-source music generator with Suno-level audio quality and sheet music editing
multimodal-art-projection/YuE
About the project
By inputting only lyrics and style instructions, it creates a sheet music plan including melody and chord progressions, and converts it into a finished track with vocals and accompaniment. The generated sheet music adopts a white-box structure that allows humans to directly view and edit, enabling configuration review or changes before the final result is produced.
It has demonstrated audio quality competitive with Suno v5/v6 in WildSongBench evaluations, showing particularly high accuracy in its zero-shot cover feature, which transcribes existing recordings and reinterprets them in new styles. A single checkpoint handles original song generation, covers, and interactive editing via agents, simplifying the workflow.
It generates 48kHz stereo audio using an AR-NAR Mixture-of-Transformers backbone and a VAE decoder. It runs on NVIDIA GPUs in BF16-supported environments, and the Python API allows step-by-step control from plan creation to semantic token generation and final decoding.
multimodal-art-projection/YuE
YuE2: frontier music generation with symbolic planning, zero-shot covers, and agentic music editing.
Python
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.

