Has anyone ever tried to transcribe audio using two or more speech to text APIs? You could match up the words using time stamps, and pick the word with the highest confidence. If anyone has done this, did it significantly improve the accuracy of the transcription? Is it worth it?