Skip to main content
LemonSlice generates real-time, interactive video avatars for voice and video agent applications — lip-synced video over WebRTC, with zero-shot avatars from any reference image. Already running a voice agent? Keep your STT, LLM, and TTS. LemonSlice adds synced avatar video on top, streamed over WebRTC into the same session. See Reference to connect it.

Quickstart (5 min)

Clone a production-ready Next.js + LiveKit agent
Each option delivers the same core avatar video. They differ in who runs the speech and intelligence stack, and how much UI you build yourself.
Self-Managed Pipeline (LiveKit, Agora, etc)WidgetHosted pipeline
⭐⭐⭐⭐⭐⭐
Coding levelFull customizationNo-code (like embedding a YouTube video)Low-code (UI customization only)
Avatar videoLip-synced realtime videoLip-synced realtime videoLip-synced realtime video
Speech & intelligenceYou choose any STT, LLM, and TTSLemonSlice-managedLemonSlice-managed
FrontendYou build your call UIYou get our prebuilt UIYou build your call UI
Avatar selectionAny image at runtimeManually design an avatar in the web appManually design an avatar in the web app
Best forProduction agents, custom stacks, full controlSites with no backendCustom UI without running your own AI stack
Add LemonSlice as the video layer on top of your agent stack — available as a LiveKit integration or Pipecat integration. You control STT, LLM, TTS, and the call experience; LemonSlice handles avatar generation only.

​
Widget

A prebuilt, embeddable web component for a no-code embed on your website. Design your agent in our web app and paste one snippet — no backend required. See the Widget overview. Best when you need a corner chat widget on a marketing or content site, not a custom agent application.

​
Hosted pipeline

LemonSlice runs STT, LLM, and TTS; you control the frontend via Daily rooms. See the Hosted pipeline overview. Consider this when you want a custom call UI but do not want to manage the AI pipeline yourself.