I've heard textual description tracks on television programs before. They come fast, but not screen-reader fast. To the untrained ear a blind person's screen reader sounds like when you somehow get the TI-99/4A's speech synthesizer to read from invalid memory.
The audio description tracks are a different genre than screenreadera perform. They're acting, by actors, carefully written and performed to fit into the gaps in the dialogue while preserving the mood and flow of the show. I think speeding them up or making them robotic would ruin them, while both of those traits are actually desirable for screenreaders.