Longtime media maven Jeremy Wagstaff, who has done stints with the WSJ and other legit media, has penned a very insightful piece about Voice AI and Transcription. In the post, Jeremy highlights services that are doing decent work in the AI-assisted transcription industry.
I read this with a keen interest, as my agency's very first technology client was a startup, MobileWord, way back in the mid-90s. A lot of what is now more than possible, and actually working, was envisioned back then by founder Bob Cox, but he was way too early, and the computing power to do what is now possible today just wasn't there. Back then, the files were shuttled to human transcribers, and then returned to the person who had done the dictation via email. It was a cutting-edge idea back then, but today seems so expected.
The big issue remains transcription of accents and regionalized dialects. This is one of the learning hurdles that the services are all just starting to address. One person's English from say Seattle is far different in accent to a Bostonian. Add in EASL (English as a Second Language) speakers from South America, Asia, Europe or Africa and the learning that the voice engines need to undertake is a massive computing job.
Today, with the kind of cloud computing power that's available, Bob's vision is now a reality. Give Jeremy's post a read and check out some of the services he highlights.