Industrial-grade speech recognition toolkit. 170x realtime, 50+ languages, speaker diarization, emotion detection — all in 3 lines of Python. Production-ready.