v0.10.0

v0.10.0 (2026-04-30)

Features

• expressions, embedder, wavtool: Add Breathiness & Voicing expression parameters, mHuBERT backend, and HPSS-based extraction (55fe4f4)

• expressions: Add BrecLoader and VoicLoader for Breathiness and Voicing expression parameter extraction using HPSS-based breath/voice index analysis (55fe4f4)

• wavtool: Add extract_wav_breath_voice() for HPSS frequency-band RMS extraction; add extract_wav_embeddings() with mHuBERT/PCA support and caching (55fe4f4)

• embedder: Introduce mHuBERT as advanced speech feature extraction backend for improved cross-language phoneme alignment (55fe4f4)

Refactoring

• expressions: Rename pitd cache dir from "pitd" to "f0"; remove add_cuda_to_path import in crepe branch; force RMVPE to device="cpu" (55fe4f4)

Documentation

• readme: Restructure ML models section with Pitch Extraction and mHuBERT Feature Extractor subsections; add CC-BY-NC-SA-4.0 license warning (55fe4f4)

• readme: Add Get Model Weights section with HF mirror tip and separate RMVPE/mHuBERT download instructions (55fe4f4)

• readme: Add Breathiness and Voicing to supported params and features list; add BrecLoader/VoicLoader subgraphs to algorithm workflow diagram; fold v0.5.0/v0.6.0 version tips into inline bullet text; remove v0.9.1 tag (55fe4f4)

• examples: Add demo scripts for mHuBERT embedder and Breathiness/Voicing expressions: mhubert_embedder_demo.py and breath_voicing_control.ipynb (55fe4f4)

Breaking Changes

• expressions: pitd cache directory renamed from "pitd" to "f0"; existing caches and scripts referencing the old path must be updated

---

Detailed Changes: v0.9.0...v0.10.0

Acknowledgements: Special thanks to @ma0shu for foundational work on expressive parameter extraction ❤️ (ma0shu@a15858d)

License Reminder: Model weights of mHuBERT are under CC-BY-NC-SA-4.0 license and not included in this release. Please review LICENSE and README.md for full terms.