The results of using a saved/trained speaker voice, and then in the editor being able to assign an image to produce an animated avatar works well.

However, it appears you always have to add the avatar image as an option in a project, and the "AI Speakers" pane doesn't let you create or edit speakers to be able to add an image.

It would be great in the "AI Speakers" pane you could create/edit the speakers, and that include assignment of the avatar image. It would also be natural to allow multiple images per ai speaker so you could have the avatar in a number of different scenarios/clothes etc with same voice.