YuE2
Open lyrics-to-song model that writes the score before it sings — full 48 kHz stereo songs, offline.
About
YuE2 is the second generation of YuE, the open lyrics-to-song foundation models from M-A-P and HKUST. Give it a style prompt and lyrics and it produces a complete song with vocals and accompaniment, with the distinctive twist that it writes an editable ABC melody-and-chords score first and then renders that score into 48 kHz stereo audio. It supports multiple genres, languages and vocal techniques, and the release ships with inference code, checkpoints, and demo pages.
The whole thing runs locally with no API needed, though full-length songs want serious VRAM (shorter clips fit in ~24 GB, long songs in ~80 GB+ per the docs). Weights are released under CC BY-NC 4.0 with an explicit carve-out letting personal users, creators and musicians monetize their outputs; companies need a commercial license. Early coverage notes its own caveat that the top-line SongBench gap to proprietary systems is not statistically significant.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
- Provider-reported SongBench scores on the model card; authors' README notes the gap at the top is not statistically significant
- Early third-party write-ups echo that caveat rather than independently re-running evals
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (TIER metrics, metrics-v1). Evidence: ~12k GitHub stars on the YuE repo (updated for YuE2); Apache 2.0 code; paper + model card with honest benchmark caveats (arXiv 2503.08638). Strengths: genuinely open weights, rare score-first workflow, active community tooling. Docked for: (1) weights under CC BY-NC 4.0 — not commercially clean; (2) benchmark evidence is author-reported with a stated significance caveat. License: Apache 2.0 (code); CC BY-NC 4.0 + creator permission (weights).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · YuE repo ~12,000 GitHub stars, updated for YuE2 (verified 2026-10-16)
- · Apache 2.0 code; YuE2-3B weights on Hugging Face (m-a-p)
- · Model card carries honest benchmark caveats (SongBench gap not significant)