myshell-ai/OpenVoice

The repository now includes the full OpenVoice V1 and V2 codebases, featuring instant voice cloning, flexible voice style control, and zero-shot cross-lingual capabilities. V2 adds native multi-lingual support for English, Spanish, French, Chinese, Japanese, and Korean, alongside improved audio quality. The project is now licensed under the MIT license, allowing free commercial and research use. The release includes demo notebooks for voice style control, cross-lingual cloning, and multi-accent generation using MeloTTS, along with a Python package setup for easy installation.

(repo-wide) · high confidence

The openvoice package has been refactored into a Python-compatible package structure, introducing new API classes (BaseSpeakerTTS, ToneColorConverter) and a Gradio-based demo application. This update adds support for English and Chinese text-to-speech, tone color conversion for voice cloning, and watermarking for generated audio. The implementation includes new modules for audio processing, model architecture, and text cleaning, enabling users to clone voices from short audio clips with granular control over style and emotion.

openvoice · high confidence

The project now includes a requirements.txt file specifying 16 Python packages, including librosa, faster-whisper, pydub, and gradio, which provide the necessary libraries for audio processing, speech recognition, and user interface components.

(dependencies) · medium confidence

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

• CAI 46 → 46 (-0.1)
• Rubric changed (rubric-2026.08.15 → rubric-2026.09.15) — scores are not directly comparable.

• Code Health 100 → 99 (-1.2)
• Architecture 94 → 97 (+2.8)
• Maturity 46 → 39 (-7.0)
• Readiness 17 → 20 (+3.3)
• Security 100 → 98 (-2.5)

• Dimension evaluation failed
• No exposed public API
• The README does not mention any training data or model size, which is critical for an AI researcher audience who may want to evaluate the model's scale. (README.md)

• Banned license: dtw-python
• Banned license: unidecode
• Critical CVE: [GHSA redacted] (requirements.txt)
• Documentation: no installation or build instructions (README.md)
• Documentation: no project overview (README.md)
• Documentation: no usage examples (docs/USAGE.md)
• Duplicated block (11 lines × 2) (openvoice/modules.py)
• Duplicated block (13–15 lines × 2) (openvoice/utils.py)
• Duplicated block (19 lines × 2) (openvoice/modules.py)
• Duplicated block (1–24 lines × 4) (openvoice/mel_processing.py)
• Duplicated block (6 lines × 2) (openvoice/attentions.py)
• Duplicated block (6 lines × 2) (openvoice/attentions.py)
• Duplicated block (6 lines × 2) (openvoice/text/mandarin.py)
• Duplicated block (8 lines × 2) (openvoice/attentions.py)
• Duplicated block (8 lines × 2) (openvoice/transforms.py)
• Orphaned files with no living knowledge
• Outdated: cn2an
• Outdated: faster-whisper
• Outdated: gradio
• Outdated: inflect
• …and 5 more

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

myshell-ai/OpenVoice was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.