funasr

github.com ↗

Industrial-grade speech recognition toolkit with 170x realtime speed, 50+ languages, speaker diarization, and emotion detection.