I need to record natural language and transcribe to text

I am looking to develop a solution which allows for a recording device such as a google home or generic recording hardware to record a natural conversatio. And transcribe into text. I would like for the recordings to be captured in separate instances as.the solution should be able to detect when conversations end and new instances begin. I would then like to be able to obtain details on each instance such as the duration of the conversation, possibly identifying attributes like gender, etc.