Lokutor Is Now a LiveKit Agents Plugin

Lokutor is now a plugin for LiveKit Agents. livekit-plugins-lokutor was merged into livekit/agents on 12 November 2026, and it gives any LiveKit agent our speech recognition and our voices.

What is LiveKit Agents?

LiveKit is open-source infrastructure for realtime audio and video. LiveKit Agents is its Python framework for voice agents that join a room or a phone call. The framework runs the conversation: the room, turn-taking, interruptions and the language model. Plugins supply the speech in and out.

What the plugin adds

• lokutor.STT: Conv 2.0, our speech recognition model. It streams interim and final transcripts as the user speaks, written out with punctuation, casing and digits. If you don't set the language, it is detected.
• lokutor.TTS: Versa 2.0, our voice model. Ten voice slots, each with a native speaker per language. The LLM's reply is spoken sentence by sentence as it is written, so the agent starts talking before the reply is finished.

Both work in English, Spanish, Catalan, Galician, Basque, Portuguese, French, Italian and German. Each direction is one model for all nine languages, so there is no per-language routing to set up.

The code

from livekit.agents import Agent, AgentServer, AgentSession, JobContext, cli
from livekit.plugins import lokutor, openai, silero

server = AgentServer()

@server.rtc_session()
async def entrypoint(ctx: JobContext) -> None:
    session = AgentSession(
        stt=lokutor.STT(language="es"),
        llm=openai.LLM(),
        tts=lokutor.TTS(voice="F1", language="es"),
        vad=silero.VAD.load(),
    )
    await session.start(agent=Agent(instructions="Eres una recepcionista amable."), room=ctx.room)

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

If you already run a LiveKit agent, switching to Lokutor's speech means changing the stt= and tts= lines; the rest of the agent stays as it is.

Install it with LiveKit Agents:

pip install "livekit-agents[lokutor]~=1.8"

The package is expected to be published to PyPI with LiveKit's next release. Until then, the integration guide shows how to install it from LiveKit's repository.

Why it matters

• Spanish, Catalan, Galician and Basque in one model. A LiveKit agent built for Spain can hear and answer in all four, each with its own native voices, alongside English, Portuguese, French, Italian and German.
• Models we build ourselves. Recognition and voice are our own models, trained and served by us, not a resold third-party API.
• Pay for what you use. The plugin is billed at the same prices as our API, from a prepaid balance, with $5 of credit when you sign up. See the pricing page.

If you want Lokutor to run the whole conversation, including the LLM, turn-taking, tools, phone numbers and recordings, use a Lokutor agent instead. The plugin is for teams who already build on LiveKit and want our speech inside their pipeline.

Get started

1. Create an API key at app.lokutor.com and set it as LOKUTOR_API_KEY.
2. Follow the LiveKit integration guide. It covers the parameters, concurrency and billing.
3. Read the source in livekit/agents.

Thanks to the LiveKit team, who reviewed the plugin closely over several rounds. If you build something with it, tell us at [email protected].

Try the voices in your browser on Hugging Face. Lokutor is live on Product Hunt.