I doubt that the google speech model was meant to deal with AMBE compressed voice. I think you will need to create your own speech-to-text model to solve the q

I doubt that the google speech model was meant to deal with AMBE compressed voice. I think you will need to create your own speech-to-text model to solve the quality issues.