Avatar Studio
Configure a GLB scene, animation states, timed media and AI-triggered actions, then test live conversations before publishing.
Open a character and choose Edit & Test in the Avatar Studio card. The character editor saves changes before opening the studio. The existing /playground/ URL remains the address of this workspace.
Arrange the scene
Use the scene hierarchy to select the environment, ground, camera, character, lights and props. Set a background or platform, frame the character and save the scene. The embedded scene templates use this saved presentation.
Choose the three animation states
| State | When it is used |
|---|---|
| Idle | The character is waiting without active listening or speech. |
| Listening | The character is receiving visitor speech. |
| Talking | The character is speaking. |
Use animation clips authored for the same or a compatible skeleton. Facial tracks also require matching morph targets. A compressed GLB is supported when the model and animation formats match the loader; compression does not fix an incompatible rig.
Build an action
- Add a rule
Open Actions and add an animation, narration or video action. Give it a useful name.
- Select a trigger
Choose an inactivity delay or an AI action code. Inactivity rules begin when the experience is ready; a visitor does not need to send the first message.
- Combine the media
Choose an animation, narration source, optional background music and optional video. Position the video screen in the scene and adjust the volumes.
- Preview and save
Use Try now to check the action, then save the scene. Configure repetition and the delay between repetitions if needed.
AI action codes
Each action can have its own code. For example, configure greeting and let the model emit [[action:greeting]] when that action is appropriate. The runtime recognizes the explicit marker; merely writing the word “greeting” in normal conversation does not select the action. Multiple configured actions can use different codes.
Narration, music and video
Use an audio file for a prepared spoken explanation, or choose text narration. Background music has its own volume. Video is displayed on the scene screen; the separately configured audio supplies the narration and soundtrack.
Uploaded narration is prepared by the speech-to-face service so its facial animation can be reused at playback. The action plays the prepared audio and facial timing together. This is an action-specific workflow; normal live conversation continues to use its own speech pipeline.
Test the complete experience
Check the first visit, a repeated action, an interrupted action, mobile audio permission and the character’s facial mapping. Save the scene, then test the selected website template in Publish. A browser may still require a tap before audible media starts.
Continue with the website embed guide or the custom avatar guide.
Maintained by IAMX. Updated .