I was imagining a talking head avatar rendered in Unreal Engine, with lip sync and facial expressions driven by a multimodal LLM that produces the speech.