Then you need to implement a voice recognition module that converts the sound bytes into printable words. Might take you a few years with a good team of developers.