Hugging Face launches Open TTS Leaderboard for scalable multilingual text-to-speech evaluation

Hugging Face introduced the Open TTS Leaderboard on Sep 30, 2026, to provide standardized, scalable evaluation for the growing ecosystem of over 8K open-source TTS models. Unlike subjective arena-based leaderboards, which struggle with scaling and voter consistency, this new leaderboard uses objective metrics—Intelligibility (WER/CER via Qwen3 ASR), Speed (RTFx and TTFA on H200 GPU/CPU), and Speaker similarity (cosine similarity of WavLM embeddings)—to evaluate models in hours instead of weeks. The leaderboard complements, but does not replace, human preference rankings.