Thanks — I hear you: Gemini TTS latency is high and you want practical ways to improve it. Here are focused steps you can try. 1. Deploy closer to Gemini endpoints • Run your agent server in a region/network near the Gemini API to reduce network RTT. 1. Enable preemptive generation • Turn on

preemptive_generation

in your AgentSession so synthesis can start before the user finishes speaking. This reduces perceived delay. 1. Use streaming / instant mode if available • Check the Gemini TTS plugin settings for any streaming or “instant” mode and enable it. Streaming reduces Time To First Byte (TTFB). 1. Monitor TTS metrics • Track TTFB and end-to-end latency via the Agents/TTS metrics so you can see where time is spent and measure improvements. 1. Tune provider settings • Try different Gemini models or request parameters (audio format, chunking) that trade off quality for speed. If you’d like, tell me: • where your agent server is hosted (region), • whether you already have

preemptive_generation

enabled, • and any TTFB or latency numbers you’re seeing. I’ll suggest the next concrete changes based on those details. Sources: How can I reduce latency in voice agents using STT, TTS and LLM? | Simple Call Answering Agent | LiveKit docs | Agent speech and audio | LiveKit docs | Hume TTS plugin guide | LiveKit docs | TTS Metrics | LiveKit docs