Hey, I’ve helped teams optimize VAPI latency and improve first-response times on OpenAI clusters. That 8–14s delay usually comes from the model warm-up, stream buffering, or how the “wait for user” logic is handled. I can help you tweak your session settings, streaming mode, and turn-taking config to make it feel instant. are you handling the speech detection locally or through VAPI’s built-in endpoint? @vhedf25