To reduce Time To First Byte (TTFB) and improve latency, you can follow these steps: • Optimize network proximity: Ensure your agent is close (in network terms) to the services it uses, such as LLM, Speech-to-Text, and Text-to-Speech services. • For Rime TTS: Enable the

reduce_latency

parameter by setting it to

True

, adjust the

speed_alpha

parameter (values lower than 1.0 result in faster speech), and optimize the sample rate to match your application needs. • For Google Speech-to-Text: Use recognizers to improve latency and set appropriate location parameters to ensure proximity to the API endpoint. • Monitor latency: Use the Agents metrics API to monitor and understand latency issues. These steps should help in reducing TTFB and improving overall latency. Sources: livekit.plugins.google API documentation | How can I reduce latency in voice pipeline agents using STT, TTS and LLM? | Rime integration guide | LiveKit Docs