v0.2.23 yue2-music-v1.2.0
huggingface/speech-to-speechSep 24, 2026 → Sep 25, 20261 day between releases
What's New
The provided release notes describe two unrelated projects, memanto and yue2-music, with no shared features.
New Features
- Instrumental music generation and covers
- Multiple input methods including text, ABC scores, and reference recordings
- GPT-6 Astra agent orchestration for workflow management
- Vocal note preservation workflow in Ins
- Local audio and prompt export functionality
Improvements
- Optimized unfiltered recall performance in memanto
- Standalone checks and setup instructions for yue2-music
- Validation tests (51 CPU tests and 35 regression tests)
- SheetSage2 processing for audio references
Bug Fixes
- Over-fetching removed from unfiltered recall queries
- Generation limit retry logic implemented
- ABC cover retry completed after hitting limits
- Vocal leakage assessment workflow added
Upgrade Notes
The older release notes describe the memanto project (migration tools), while the newer release describes the yue2-music project (audio generation). These are different repositories with no upgrade path between them.
The Two Releases
v0.2.23
Sep 24, 2026· 28 days ago
This release introduces new migration sources for Zep and Hindsight, adds community migration showcases, and optimizes recall queries to prevent over-fetching.
Full release notesYuE2 Music skill 1.2.0: instrumental generation and covers
Sep 25, 2026· 27 days ago
This release adds instrumental generation and instrumental covers to the yue2-music skill, enabling users to create audio from text descriptions, ABC scores, or reference recordings. The workflow utilizes the GPT-6 Astra agent for orchestration and checks results.
Full release notes