Hugging Face released a scalable evaluation framework to benchmark multilingual text-to-speech and voice cloning. The leaderboard tracks model performance across diverse languages and speaker identities. It replaces subjective listening tests with standardized metrics.