Add an interactive avatar to any voice stack

LemonSlice is the face layer for voice agents. It works with any voice agent platform, any LLM or realtime model, and any TTS or STT provider, and ships live WebRTC integrations with LiveKit, Daily (Pipecat), and Agora. Your stack handles the conversation; LemonSlice streams the avatar.

Browse integrations

Every provider LemonSlice works with, in one directory. Pick a layer of the stack to see its integrations.

LemonSlice never touches your input pipeline. Any STT and VAD stack works unchanged.

OpenAI Realtime

Pair OpenAI's Realtime API, TTS, or Whisper with a real-time LemonSlice avatar for face-to-face conversation.

Gemini Realtime

Give your Gemini Live, TTS, or STT pipeline a lifelike interactive AI avatar for face-to-face conversations.

Cartesia

Stream Cartesia's low-latency TTS into the avatar for fully real-time conversations.

Deepgram (TTS)

Stream Deepgram Aura audio into a LemonSlice interactive avatar for lip-synced, face-to-face conversations.

Azure Speech

Pair Azure AI Speech voices, transcription, and realtime voice with a real-time LemonSlice avatar.

AssemblyAI

AssemblyAI transcribes what your users say, and a real-time LemonSlice avatar speaks the answer face to face.

Gradium

Gradium's streaming speech APIs pair with a real-time LemonSlice avatar for natural, face-to-face voice conversations.

Simplismart

Stream audio from models served on Simplismart into a lifelike real-time avatar for face-to-face conversation.

Speechmatics

Speechmatics multilingual speech-to-text and Flow agents paired with a real-time LemonSlice avatar your users can see.

Sarvam

Sarvam drives the Indic-language voice and LemonSlice adds the lifelike real-time avatar that speaks it.

Mistral AI

LemonSlice adds a real-time avatar to Mistral models and Voxtral speech recognition.

Baseten

Serve open speech models on Baseten and give them a real-time face with a LemonSlice interactive avatar.

NVIDIA

Pair NVIDIA Riva and NIM speech pipelines with a real-time LemonSlice avatar that speaks every response.

Don't see your provider?

If it streams audio, it works with LemonSlice. The avatar API accepts any audio source and returns a real-time video stream. No named integration required.

Your agent talks.
LemonSlice gives it a face.

LemonSlice sits on top of the voice pipeline, not inside it. Nothing about your existing agent changes.

  1. Step 1: Keep your voice agent

    Your platform keeps handling speech-to-text, the LLM, and text-to-speech exactly as it does today.

  2. Step 2: Stream audio to LemonSlice

    Pipe your agent's audio output to the LemonSlice API. The avatar renders speech, expression, and lip sync in real time.

  3. Step 3: Embed the video stream

    The avatar streams back over WebRTC into your app or site, natively through LiveKit, Daily (Pipecat), or Agora.

Frequently asked questions

Stream your agent's audio output to the LemonSlice API. The avatar renders and lip-syncs in real time and streams back over WebRTC. Your Vapi or Retell configuration stays unchanged.
LemonSlice is the face layer on top of your stack. Your platform keeps handling STT, the LLM, and TTS.
LiveKit, Daily (Pipecat), and Agora, the three WebRTC orchestration options, are live integration partners. Everything else works through the avatar API.
Yes. LemonSlice lip-syncs to any TTS audio, so your existing voices work without migration.
Yes. Speech-to-speech realtime models stream directly into the avatar.
About the blink of an eye. LemonSlice has the fastest response time of any major avatar provider, optimized for real-time conversation.

Put a face on your voice agent

5-minute API setup. Works with the stack you already have.