It is interesting. I would suggest improving it a little further:

Add a configurable voice synthesizer to read everything that arrives in conversations. You could configure what engine to use (espeak, for example) and how it would be sued: speaker, bluetooth, what contacts, and so on.

@lewwy
By the way, create a Brainstorm and put a link to this post there. Then edit you first post and put a link to the brainstorm here.