Skip to main content
If you already have a voice agent, let it continue to decide when to reply, what to say, and how to generate the voice. Vivix uses that live audio to render the avatar. For example, when a customer asks about a product, your agent produces the answer audio and the avatar speaks along with it. A standard avatar conversation does not need pure rendering mode.

​
1. Decide who generates the voice

Your application supplies live audio in this mode. For a complete prerecorded clip, use an audio script. If you want Vivix to process user input and generate the reply, use a standard session.

​
2. Create a pure rendering session

Add this configuration to a working Create Session request. Keep your first image and output settings.

​
3. Connect the live audio

Use the session credentials to join the RTC room, publish your audio track, and subscribe to the avatar video. Start with a microphone to verify the connection, then connect the audio output of your voice agent. See the TRTC or Agora integration guide.

​
4. Control the picture and close the session

The published audio determines the spoken content and voice. TTS settings and spoken instructions do not rewrite it. Use an action script for a specific movement, or switch the image to change the look. Stop publishing audio when you want to stop the input. Close the Session when the performance is over.