Tag
The author argues that for live voice agents, STT latency and real-time behavior are more critical than raw transcription accuracy, and proposes a different evaluation scorecard.