If you want your agent to chat naturally in multiple languages, a great way is to use the

VoicePipelineAgent

along with a multilingual turn detection model and clear language switching. Here’s a simple rundown: • Use the

VoicePipelineAgent

to bring together STT (speech-to-text), LLM (language model), and TTS (text-to-speech) so your conversation flows smoothly. • Set up language switching by changing the language settings on the fly for both STT and TTS parts. • Tap into the

MultilingualModel

from LiveKit’s turn detector plugin to easily handle multiple languages. This way, your agent can naturally switch languages and sound really fluent. If you want, I’m here to help with more details or guide you through the setup! Sources: complex-agents/turn-taking/README.md | examples/voice_agents/preemptive_generation.py | AI voice agents | LiveKit Docs