How far fetchedis the idea that facial expressions on the model could be manipulated to look negative/positive when the user is talking about something the soft

How far fetchedis the idea that facial expressions on the model could be manipulated to look negative/positive when the user is talking about something the software creators have different opinions on?