Note that if you do not want to use APIs, and directly perform inference on machine learning models instead, then definitely check this tutorial, in which I'll show you how you can use the current state-of-the-art machine learning model to perform speech recognition in Python.