Dynamic type

Google voice transcription is good enough that I've started using it to compose my texts.  While I speak, my voice is mostly flat and without affect because I know that it will be flattened into text to be sent.  But what if it weren't?  Type designers use leaning, kerning,  weight,  serifs, and  ligature to communicate a feeling: what if fonts could instead dynamically communicate meta-information?

This wouldn't have to be limited to emotional affect: difficulty can be used to slow down readers, weight could either indicate passages which were previously read or be used to identify passages which the reader hadn't read, and orthographic ligature could identify connection and speed of transcription.  

This is clearly not "just a feature": it would require integration with eye tracking, affective mechanisms, microphones that heard typing, and would require editing tools as well.  But the potential benefit is enormous: for a business, supporting both sides of this communication would offer an iMessage-like walled garden.  For individuals, better connection and clearer affective feedback would  enrich   relationships.