Thanks — I’m sorry you’re seeing ElevenLabs speak only the first sentence. I’m responding about that specific problem and how to diagnose/fix it. Likely cause • The TTS pipeline is chunking text for incremental synthesis. If text isn’t properly segmented (missing sentence delimiters) or the sentence-tokenizer is bypassed/misconfigured, only the first chunk will be synthesized and played. What you can try now • Ensure the text you send contains proper sentence delimiters (., !, ?). • Send each sentence as its own TTS request or explicitly split the text before sending. • Check your agent pipeline config and confirm the

tts_node

(sentence tokenizer) is enabled and not bypassed. • If you stream text incrementally, make sure each chunk is pushed to ElevenLabs (don’t buffer everything into one unbroken chunk). What I need if the problem continues • Are you using ElevenLabs via the LiveKit Agents plugin or calling ElevenLabs directly with an API key? • The agent configuration (TTS/LLM settings) or the

tts_node

config. • A short code snippet showing how you send text to TTS and how audio is played back. • Any error logs or console output when playback stops. Send those details and I’ll walk through the exact fix. Sources: examples/voice_agents | livekit/livekit_composite | DeepWiki | Pipeline nodes and hooks | LiveKit docs