Paper and speaking based programming?

Has anyone tried creating a paper or speaking based programming environment? Basically - handwriting/diagram -> text code or - audio speech -> text code The above could be done solely with machine learning or a combination of machine learning and crowdsourcing (mech turk).