1. Full-Song Generation: YuE can generate entire songs—up to five minutes long—that include both vocal and instrumental components.
2. Lyrics-Conditioned Creation: It transforms provided lyrics into cohesive musical pieces, maintaining structure and narrative throughout the song.
3. Synchronized Vocal & Instrumental Modeling: A dual-token technique ensures that vocal melodies and accompanying instruments remain in sync.
4. Advanced Audio Tokenization: A semantically enhanced audio tokenizer streamlines training and boosts generation quality.
5. Progressive Generation Approach: Using a lyrics-chain-of-thought strategy, YuE progressively builds music that adheres closely to the lyrical input.
6. Diverse Genre & Language Support: From metal and jazz to pop, rap, and more, YuE supports a wide range of genres as well as multiple languages including English, Chinese, Japanese, and Korean.
7. Customizable Prompts: Users can enrich their prompts with details—such as specific instruments, mood, vocal timbre, and vocal gender—to tailor the music to their creative vision.
8. Free & Open-Source Foundation: As a completely free project under the open-source Foundation, YuE offers cost-free access to cutting-edge AI music generation while actively inviting community contributions, collaboration, and experimentation.
