little-bear-29830
02/18/2026, 5:22 AMtall-airline-16756
02/18/2026, 5:22 AMtall-airline-16756
02/18/2026, 5:23 AMeleven_multilingual_v2. You can use ElevenLabs through:
◦ LiveKit Inference (LiveKit handles integration/billing), or
◦ The ElevenLabs plugin available for Python and Node.js (livekit-agents[elevenlabs]) so you can use your own ElevenLabs API key.
• STT: Real-time transcription is supported (e.g., scribe_v2_realtime) with automatic language detection for many languages. This is available via LiveKit Inference or the ElevenLabs STT plugin.
• How it fits in an agent: Typical setup uses an AgentSession with an STT component (to transcribe user speech) and a TTS component (to generate spoken responses). Example (Python):
session = AgentSession(
stt=elevenlabs.STT(model="scribe_v2_realtime"),
tts=elevenlabs.TTS(voice_id="YOUR_VOICE_ID", model="eleven_multilingual_v2")
)
• Authentication: You’ll need an ElevenLabs API key (set as ELEVEN_API_KEY when using the plugin).
• Deployment / Pricing notes: On LiveKit Cloud, active agent sessions are billed (agent session minutes). LiveKit Cloud also provides observability (session analytics, logs, latency) and supports deployment features like warm agents and graceful draining.
If you'd like, I can:
• Walk you through a minimal Node or Python example to get an agent running with ElevenLabs TTS/STT.
• Help you pick the right ElevenLabs model/voice for your use case.
• Explain differences between using LiveKit Inference vs. the plugin (for example, whether you want LiveKit to handle billing/integration or prefer to use your own ElevenLabs account).
Which would you like next — a short Node sample, Python sample, or help choosing models/voices?
Sources: ElevenLabs TTS | LiveKit Documentation | ElevenLabs TTS plugin guide | LiveKit Documentation | ElevenLabs STT | LiveKit Documentation | ElevenLabs STT plugin guide | LiveKit Documentation | Deploy and scale agents on LiveKit Cloud