Invitation to try RVCBench datasets and pip evaluation tools with OpenVoice

[Nanboy-Ronan]

Hi OpenVoice team,

Thank you for developing OpenVoice and sharing your models and code with the community!

I'm Ruinan, an author of RVCBench. We'd love to invite you to try our dataset with OpenVoice. Alongside conventional voice-cloning utility evaluation, RVCBench brings several additional dimensions into one benchmark:

• Utility: speaker similarity, content accuracy, speech quality, intelligibility and emotion consistency.
• Reference robustness: background noise, overlapping speakers, compression and anti-cloning protection.
• Generation robustness: long-form speech and unusual text.
• Coverage across speakers and languages: speaker demographics, multilingual and cross-lingual cloning.
• Automated evaluation via pip: pip install "rvcbench[eval]" — seven automatic metrics, prompt export and reports; also usable on your own audio.

Hugging Face dataset · Quick start · PyPI

There's a small 52-utterance onboarding suite to get started (packaged suites are currently previews). We'd be happy to help with setup or evaluating a relevant subset, and would really appreciate your feedback.

Thank you for considering it!
Ruinan, on behalf of the RVCBench team