LemonSlice x LiveKit

LemonSlice now integrates directly with LiveKit, making our real-time avatars available as a modular component that developers can drop into their own AI applications. Instead of adopting a fixed, end-to-end stack, you can bring your own AI models and infrastructure while using LemonSlice for expressive, low-latency face-to-face interaction.
Build the voice pipeline that works best for your application:
- Voice Activity Detection (VAD) & Speech-to-Text (STT): Use Silero VAD, Deepgram, AssemblyAI, OpenAI, Google, Azure Speech, or any provider supported by LiveKit Agents.
- Large Language Models (LLMs): Connect OpenAI, Anthropic, Google Gemini, xAI, Mistral, Groq, Together AI, or any OpenAI-compatible endpoint - including self-hosted models.
- Text-to-Speech (TTS): Generate speech with ElevenLabs, Cartesia, OpenAI, Deepgram, Google, Azure Speech, Hume, or dozens of other supported providers.
LemonSlice plugs directly into your existing LiveKit Agents pipeline, so you can mix and match the best models for each stage of the stack without changing your avatar layer. Whether you're optimizing for latency, quality, cost, or custom infrastructure, LemonSlice fits into your architecture - not the other way around.
The result is a composable, production-ready platform for building conversational AI with expressive, real-time avatars.



































































































































































































