For developers & agencies

Avatar API

Generate videos from text or your own recording, or give your voice agent a face. Connect with LiveKit, your app or your automation workflow.

Choose the output your app needs.

Generated video

A standard POST /inference takes text or an uploaded recording and returns JSON with video URLs for playback and download.

Streaming

The public API's ?stream=true returns MP4 data, not JSON. WebSocket endpoints support ongoing avatar streams.

Lip-sync to your own recording

curl -X POST https://api.avatartalk.ai/inference \
  -H "Authorization: Bearer $KEY" \
  -F "audio=@speech.wav" \
  -F "avatar=european_woman" \
  -F "emotion=neutral"

WAV, MP3, M4A or OGG up to 5 minutes. Each started second of the recording costs one credit. The playground generates video files. Streaming uses the public API or LiveKit integration, not the panel's /api/inference endpoint. See the docs for supported avatars and languages.

Plug into the tools you already use.

Make your first avatar video.

Try it in the browser. Connect your code when you are ready.

Open the playground