Add a face to your Mistral pipeline

LemonSlice puts a lifelike, real-time avatar on top of your Mistral pipeline. Whether you run open-weight Mistral LLMs, Voxtral speech recognition, or your own TTS, the avatar turns that audio into natural, face-to-face conversation.

LemonSlice

At a glance

Categories
Realtime, Transcription (STT + VAD)
Status
Compatible (works with any)
Type
Avatar API
Languages
Python, Node.js
Setup time
~5 minutes

How the integration works

Mistral AI keeps doing what it does today. LemonSlice turns the audio into a face.

  1. Keep your Mistral setup

    Your pipeline keeps running as is, Voxtral transcribes user speech, your Mistral model generates the response, and your TTS layer turns it into audio. Nothing in that chain changes.

  2. Stream the audio to LemonSlice

    Send the synthesized audio from your Mistral-driven pipeline to the LemonSlice avatar API. The avatar renders in real time with lip sync and expression matched to the speech.

  3. The avatar video streams to your users

    LemonSlice delivers the avatar video over WebRTC, natively via LiveKit, Daily (Pipecat), or Agora. Your users see a responsive face in about the blink of an eye.

Add your interactive avatar in less than 5 minutes

A Mistral AI session with a LemonSlice avatar.

Choose a code example
from livekit import agents
from livekit.agents import AgentSession, RoomOutputOptions
from livekit.plugins import lemonslice

async def entrypoint(ctx: agents.JobContext):
    await ctx.connect()

    session = AgentSession(
        # Add Mistral AI STT, plus your LLM and TTS, here
    )

    avatar = lemonslice.AvatarSession(
        # Publicly accessible image URL for the avatar
        agent_image_url="...",
        # Prompt to guide the avatar's demeanor while speaking
        agent_prompt="a person talking",
        # Recommended. Adapts video quality to each viewer's bandwidth
        simulcast=True,
    )

    # Start the avatar and wait for it to join
    session_id = await avatar.start(session, room=ctx.room)

    # Start your agent session with the user
    await session.start(
        # ... room, agent, room_options, etc....
    )

An interactive AI avatar that feels alive

Control actions and emotions

The Action Engine animates the whole performance, face, hands, shoulders, and posture. You trigger actions and emotions from the API, down to a prompt like "talk with hand gestures."

Works with any character, in any style

LemonSlice stands alone in animating non-humans, a turtle or a smiling teacup, with the same ease as cartoons and photorealistic people.

Ultra-low latency

LemonSlice delivers the quickest response time among major avatar providers, so replies feel instant.

Create an avatar from one image

An interactive avatar starts from one photo. No training or fine-tuning needed. Make unlimited custom characters and change outfits or settings in real time.

Built for enterprise scale

Avatar sessions alongside Mistral AI run at production scale.

Thousands of concurrent calls

Run thousands of avatar sessions side by side, ready for high-volume production workloads.

Multi-region servers

Launch avatar sessions from the region closest to your users so latency stays minimal.

Zero data retention mode

Run avatar sessions that store no conversation data, made for privacy-sensitive enterprise workflows.

What builders ship with Mistral AI + LemonSlice

Create avatars for everything from kids tutoring to interviewing, corporate training, and customer support.

Kids tutor

A patient tutor that answers every why.

Interviewer

Structured screening interviews, face to face.

Corporate training coach

Roleplay difficult conversations and train staff on policy.

Frequently asked questions

Keep your Mistral models and Voxtral transcription in place, then stream your pipeline's synthesized audio to the LemonSlice avatar API. The avatar video streams back to your users over WebRTC.
No. Mistral handles the intelligence across transcription and conversation. LemonSlice handles the face with real-time avatar rendering and video delivery.
Yes. LemonSlice only consumes the audio your pipeline produces, so self-hosted open-weight models work the same as hosted ones.
Yes. Voxtral keeps transcribing user speech on your side, and the avatar lip syncs to whatever audio your pipeline sends.
Open-weight Mistral LLMs drive the conversation and Voxtral handles speech recognition, while LemonSlice renders the real-time avatar from the audio your pipeline sends.
No. Your models, prompts, and transcription setup stay exactly as they are, since LemonSlice only renders the face from your pipeline's audio.

Give your Mistral agent a face

Keep your pipeline, add a real-time avatar with a 5-minute API setup.