Not a common use case, but if you want to initiate speech before the avatar is visible and display the avatar mid-speech, or hide the avatar mid-speech, that can easily be done by setting Scene dimensions to 0,0. Re-display the avatar by using dynamicResize() to set non zero dimensions.

Note that this is very different from using loadSceneByID() & unloadScene(). When Scene is unloaded, api calls cannot be used, and speech is halted. In contrast, the method we demonstrate here keeps the Scene loaded at all times.