Pipeline Translator Agent | LiveKit Documentation

This example shows how to build a simple voice-to-voice translator: listen in English, translate with an LLM, and speak the result in French with ElevenLabs TTS. Instead of using LiveKit Inference, this example uses agent plugins to connect directly to OpenAI and ElevenLabs.

Prerequisites

Add a .env.local in this directory with your credentials:

LIVEKIT_URL=your_livekit_url
LIVEKIT_API_KEY=your_api_key
LIVEKIT_API_SECRET=your_api_secret
OPENAI_API_KEY=your_api_key
ELEVENLABS_API_KEY=your_api_key
DEEPGRAM_API_KEY=your_api_key

Install dependencies:

pip install "livekit-agents[openai,elevenlabs,deepgram]" python-dotenv

Load environment, logging, and define an AgentServer

Load your .env.local and set up logging to trace translation events.

import logging
from dotenv import load_dotenv
from livekit.agents import JobContext, AgentServer, cli, Agent, AgentSession
from livekit.plugins import openai, deepgram, elevenlabs

load_dotenv(".env.local")

logger = logging.getLogger("pipeline-translator")
logger.setLevel(logging.INFO)

server = AgentServer()

Define the translation agent

Keep the agent lightweight with focused instructions: always translate from English to French and respond only with the translation.

import logging
from dotenv import load_dotenv
from livekit.agents import JobContext, AgentServer, cli, Agent, AgentSession
from livekit.plugins import openai, deepgram, elevenlabs

load_dotenv(".env.local")

logger = logging.getLogger("pipeline-translator")
logger.setLevel(logging.INFO)

server = AgentServer()

class TranslatorAgent(Agent):
    def __init__(self) -> None:
        super().__init__(
            instructions="""
                You are a translator. You translate the user's speech from English to French.
                Every message you receive, translate it directly into French.
                Do not respond with anything else but the translation.
            """
        )

    async def on_enter(self):
        self.session.generate_reply()

Define the rtc session with translation pipeline

Create the session with Deepgram STT, OpenAI LLM, and ElevenLabs multilingual TTS for French output.

import logging
from dotenv import load_dotenv
from livekit.agents import JobContext, AgentServer, cli, Agent, AgentSession
from livekit.plugins import openai, deepgram, elevenlabs

load_dotenv(".env.local")

logger = logging.getLogger("pipeline-translator")
logger.setLevel(logging.INFO)

server = AgentServer()


class TranslatorAgent(Agent):
    def __init__(self) -> None:
        super().__init__(
            instructions="""
                You are a translator. You translate the user's speech from English to French.
                Every message you receive, translate it directly into French.
                Do not respond with anything else but the translation.
            """
        )

    async def on_enter(self):
        self.session.generate_reply()

@server.rtc_session(agent_name="my-agent")
async def entrypoint(ctx: JobContext):
    ctx.log_context_fields = {"room": ctx.room.name}

    session = AgentSession(
        stt=deepgram.STT(),
        llm=openai.responses.LLM(),
        tts=elevenlabs.TTS(model="eleven_multilingual_v2"),
        preemptive_generation=True,
    )

    await session.start(agent=TranslatorAgent(), room=ctx.room)
    await ctx.connect()

Run the server

Start the agent server with the CLI runner.

import logging
from dotenv import load_dotenv
from livekit.agents import JobContext, AgentServer, cli, Agent, AgentSession
from livekit.plugins import openai, deepgram, elevenlabs

load_dotenv(".env.local")

logger = logging.getLogger("pipeline-translator")
logger.setLevel(logging.INFO)

server = AgentServer()


class TranslatorAgent(Agent):
    def __init__(self) -> None:
        super().__init__(
            instructions="""
                You are a translator. You translate the user's speech from English to French.
                Every message you receive, translate it directly into French.
                Do not respond with anything else but the translation.
            """
        )

    async def on_enter(self):
        self.session.generate_reply()


@server.rtc_session(agent_name="my-agent")
async def entrypoint(ctx: JobContext):
    ctx.log_context_fields = {"room": ctx.room.name}

    session = AgentSession(
        stt=deepgram.STT(),
        llm=openai.responses.LLM(),
        tts=elevenlabs.TTS(model="eleven_multilingual_v2"),
        preemptive_generation=True,
    )

    await session.start(agent=TranslatorAgent(), room=ctx.room)
    await ctx.connect()

if __name__ == "__main__":
    cli.run_app(server)

Run it

lk agent console pipeline_translator.py

How it works

Deepgram handles English speech-to-text transcription.
OpenAI generates a French translation from the transcript.
ElevenLabs multilingual TTS speaks the translated text in French.
The agent triggers an initial response on entry so the user hears French output immediately.

Full example

import logging
from dotenv import load_dotenv
from livekit.agents import JobContext, AgentServer, cli, Agent, AgentSession
from livekit.plugins import openai, deepgram, elevenlabs

load_dotenv(".env.local")

logger = logging.getLogger("pipeline-translator")
logger.setLevel(logging.INFO)

class TranslatorAgent(Agent):
    def __init__(self) -> None:
        super().__init__(
            instructions="""
                You are a translator. You translate the user's speech from English to French.
                Every message you receive, translate it directly into French.
                Do not respond with anything else but the translation.
            """
        )

    async def on_enter(self):
        self.session.generate_reply()

server = AgentServer()

@server.rtc_session(agent_name="my-agent")
async def entrypoint(ctx: JobContext):
    ctx.log_context_fields = {"room": ctx.room.name}

    session = AgentSession(
        stt=deepgram.STT(),
        llm=openai.responses.LLM(),
        tts=elevenlabs.TTS(model="eleven_multilingual_v2"),
        preemptive_generation=True,
    )

    await session.start(agent=TranslatorAgent(), room=ctx.room)
    await ctx.connect()

if __name__ == "__main__":
    cli.run_app(server)