YuE · s1-7B / s2-1B
Turn lyrics and style descriptions into songs with vocals and accompaniment.
GenerationOpen-source code7B + 1B
- Start here
- Test a short verse and chorus with your own lyrics. Listen for diction, voice consistency and whether the accompaniment supports the phrasing.
- Training data
- The linked release card does not provide an itemized training collection; the technical report describes the research setup.
- Training rights
- No track-level rights ledger is supplied in the card. Reference audio, lyrics and recognizable voices need their own permission assessment.
- License & output terms
- Model and weights: Apache 2.0, per the current release card.
- Limitations
- A multi-stage, memory-intensive workflow. Lyric alignment, originality and musical continuity need listening review.
- Technique
- Two-stage audio language model
- Runtime
- Local GPU inference; quantized community implementations have different hardware requirements.
- Parameters
- 7B + 1B
Sources reviewed . Training and rights notes summarize provider disclosures. A model license does not clear every source recording or generated output. Parameter counts refer to the named variant or component, not total runtime memory.