Freely available speech to text is garbage compared to what the big players are doing. Additionally their activation words usually trigger a special microcontroller extensively trained to listen for such a phrase, and then the more expensive processing kicks in.