Microsoft has published a patent application for an AI system capable of modifying speech content in real time during live broadcasts while synchronously adjusting the speaker's mouth and facial movements—so that viewers see a natural-looking modified video.