You’re probably never going to get a decent transcript from a speech-to-text engine unless you have one person speaking at a time in a crystal clear file. Your best bet is to go for the human solution (for now!)