Unverified60% confidenceBenchmarkExact time
In the end-to-end latency of voice agents, STT typically takes 200–500 ms.
2
Sources
60%
Confidence
Long-term
Relevance
7/13/2026
First Seen
Sources
状态机语音Agent实战：Pipecat与Vapi的延迟与准确性权衡
redditr/aiagents7/12/2026
Related Claims
Unverified业界领先的STT实时转录延迟约在200-400毫秒，采用Whisper或DeepSpeech等模型82% similarVerified800毫秒是语音代理端到端延迟的业界公认阈值，人类面对面对话回应延迟通常在200-500毫秒之间75% similarUnverified三明治架构相比原生Speech-to-Speech多了两次转换环节，STT环节通常增加100-300毫秒，TTS环节增加100-200毫秒74% similarUnverifiedTraditional TTS solutions require generating a complete audio file before playback, with latency typically ranging from several seconds to over ten seconds.73% similarUnverifiedA streaming pipeline can compress end-to-end latency for voice AI to under 500 milliseconds, whereas the total latency of traditional batch processing can be as high as 3–5 seconds.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/502109API
curl https://kongchang.com/api/v1/knowledge/claims/502109MCP
get_claim(id=502109)