Skip to main content
Integration Diagram LemonSlice adds a video layer on top of your agent stack. You bring your own STT, LLM, and TTS. LemonSlice listens to your agent’s audio and streams lip-synced avatar video back into the session over WebRTC. Connect with LiveKit or Pipecat:
The speed of your STT, LLM, and TTS processing directly affects the avatar’s response time and interruption handling. Optimize these components for low latency to ensure natural, real-time conversations with timely interruption handling.

​
How to use

  1. Pick a framework — follow the LiveKit integration or Pipecat integration guide.
  2. Build your UI — run your own call lifecycle and frontend around the avatar session. See Best practices for call lifecycle patterns.
You can create an agent in the LemonSlice web app and reuse its agent_id, but this is optional. You can also pass any reference image at session creation time.

​
Pricing

View pricing for video-only sessions on the LemonSlice pricing page.