From the source
Hugging Face introduced the Open TTS Leaderboard, which uses objective metrics—intelligibility (WER/CER), speed (RTFx, TTFA), and speaker similarity (SIM)—to evaluate multilingual text-to-speech and voice cloning models.
- The leaderboard aims to scale evaluation from weeks to hours, complementing human-preference-based arenas.
- It supports multilingual and voice cloning comparisons, and includes a "Listen" tab for exploring model outputs and a "Streaming" tab for streaming performance.






