Early builds of the MRTK supported Speech to Text but not the other way around. Many developers wanted a way to use Text to Speech, so Jared created a component and contributed it to the toolkit. This component is multi-threaded, supports multiple voices, integrates with Unity’s audio engine, and supports spatialization. It still exists in MRTK today. (Article, Source)

Keynote Build 2017
Several projects and hard work showcased during Alex Kipman’s keynote at the //Build conference.
