Conversation logic
Your LiveKit agent continues to handle STT, LLM, TTS, tools, and turn-taking.
YOUR EXISTING LIVEKIT AGENTGive your LiveKit agent a real-time face.
Keep your voice stack. Add synchronized motion and an on-device 3D avatar with the official Spatius plugin.
KEEP YOUR LLMKEEP YOUR TTSRENDER LOCALLY
Run Quickstart.
SpatiusSPATIUS_API_KEY · SPATIUS_APP_ID · SPATIUS_AVATAR_ID
LiveKit CloudLIVEKIT_URL · LIVEKIT_API_KEY · LIVEKIT_API_SECRET
Google AI StudioGOOGLE_API_KEY · GEMINI LIVE API ACCESSspatius@livekit — bash
Add a face. Keep the brain.
Your existing LiveKit agent stays in control. Spatius adds synchronized motion, LiveKit carries it, and AvatarKit renders the final avatar on the client.
Synchronized motion
The official Spatius plugin turns agent speech into real-time facial motion.
OFFICIAL SPATIUS PLUGINAudio + motion
LiveKit transports the audio and compact motion data together in the same room.
LIVEKIT ROOM
Final pixels, locally
AvatarKit RTC renders the avatar on the client instead of streaming cloud-rendered video.
AVATARKIT RTC · ON DEVICESpatius publishes synchronized audio and motion data—not finished avatar video. The Web client must use AvatarKit RTC.
Motion travels. Pixels stay local.
LiveKit transports synchronized audio and motion. AvatarKit turns that data into a transparent, interactive 3D avatar on the user’s device.
01 · LIVEKIT ROOM
Audio and motion, kept in sync.
LiveKit carries the agent voice and Spatius motion data together.
02 · BROWSER
Pixels rendered in the browser.
AvatarKit RTC renders the avatar locally.
Lightweight transport
LiveKit carries synchronized audio and motion data instead of a continuous rendered video stream.
Interactive by design
The transparent 3DGS avatar layer can sit naturally inside your interface, foreground, and background.
Scales with the client
Rendering runs in the browser, reducing the server-side GPU load associated with per-session video rendering.
Economical at runtime
A motion-first architecture avoids rendering and streaming a unique video feed for every session.
SayMei made AI Mandarin practice feel more face-to-face with Spatius.
LiveKit keeps handling the real-time voice session. Spatius adds the visible avatar layer that makes language practice feel more face-to-face.
Read the customer story
LiveKit integration questions.
Four practical answers, ordered from architecture risk to implementation detail.
Does Spatius replace my LiveKit voice stack?+
No. Your VAD, STT, LLM, TTS, tools, and business logic stay in your LiveKit Agent. Spatius adds synchronized motion and on-device rendering.
Why does a normal LiveKit video renderer show a black frame?+
The avatar track carries synchronized motion data, not finished video frames. AvatarKit RTC decodes the track and renders the avatar locally.
Which platforms are supported?+
The LiveKit Agents integration currently targets Web. Use the Spatius native SDKs for iOS and Android through their supported integration paths.
Where should credentials live?+
Keep Spatius and LiveKit secrets on the server. Never embed SPATIUS_API_KEY or LIVEKIT_API_SECRET in browser code.
Your LiveKit agent already has a voice.
Give it a face.
Add a synchronized, locally rendered avatar in minutes.
Start Free