Skip to main content

LiveSpeak

Real‑time voice & lip‑sync layer for 3D avatars
Turn any MetaPerson avatar into a responsive, speaking character inside your app. Supports text‑to‑speech (Azure, ElevenLabs), ChatGPT conversations, and raw audio lipsync.

Core Features​

  • MetaPerson avatar – load a GLB model of the avatar created with MetaPerson Creator and exported with Visemes.
  • Text‑to‑speech engines – supports Azure Speech and ElevenLabs.
  • ChatGPT chatbot mode – ask questions, get AI‑generated replies, spoken by the avatar with lipsync.
  • Real‑time local lipsync – feed raw audio samples (PCM, microphone, or stream) – phonemes are mapped to facial blendshapes directly in the browser.
  • Iframe embeddable – embed the tool into your app via <iframe>.
  • JavaScript API – manage LiveSpeak with JS API.

How It Works​

  1. Specify a URL of the avatar (GLB with visemes) or use the default model.
  2. Provide a text (or let ChatGPT generate a response).
  3. LiveSpeak sends the text to a TTS service and receives an audio stream.
  4. The avatar’s facial blendshapes are animated in real time to match the audio.
  5. For voice‑only input (microphone or raw audio), the built‑in local lipsync engine analyzes the waveform and drives the avatar without external TTS.

Support​

Have a question? Contact us at support@avatarsdk.com.