I wish my company's SR technology was a fraction as good as my transcription company, Nuance Communications (NCS), represents it to be to physicians and hospitals, who never see the raw output. I worked with another SR engine made by NCS's chief rival, M*Modal, which was truly superior. It had built-in productivity tools for us editors. Besides learning speakers' voices, it learned on the back end as well, meaning that it picked up on an individual editor's style and adjusted accordingly. The potential is there.

Dragon Naturally Speaking is the flagship consumer product of NCS, who are also behind the embedded SR technology in the iPhone and elsewhere. (AFAIK they are not behind the Siri speech engine; speech synthesis is a whole 'nother deal.) The draft text from our commercial product can be so bad that a handful of editors actually buy and install Dragon and redictate what they hear rather than attempting to edit the draft version, believing redictation to be faster. It's also used as a way to compensate for or prevent carpal tunnel syndrome and other RSIs.

"My primary musical instrument is the personal computer."