We have been writing this blog for quite a few years now, our first post was back in early 2009. Since then we have covered advances in the voice recording world and touched on speech recognition software here and there. We were even resellers for Dragon NaturallySpeaking for Windows for a while and were helping MacSpeech in the early days of voice recognition for Mac as beta testers. Of late we have moved away from selling voice recognition software but have always kept a keen eye on the advances.
What we typically saw with Nuance were year on year releases for their Windows product, Dragon NaturallySpeaking, each year claiming better accuracy and faster transcription but nothing to really shout from the rooftops about. On the Mac side, we saw Nuance take over from MacSpeech and redevelop an already good product into the equivalent of the Windows version feature wise.
Then, something happened with the release of Dragon for Mac v6. It felt different from the start, the voice training was minimal. The real time voice to text was spot on but what really caught my attention was the transcription.
We have been banging the drum for a while about using Dragon to speed up your transcription time and reduce your transcription costs, have a look at our post from back in 2013 – Transcription Typists, Tame The Dragon Don’t Fear It
With our testing of the latest version of Dragon (we use Dragon for Mac but Dragon for Windows has the same functionality) we have found considerable quality improvements transcribing both single speaker and multi speaker audio. This is a big deal, multi speaker audio has always been an issue with earlier versions of Dragon.
On top of that, transcribing a voice without the need to train a profile is very possible. Meaning, the audio you receive to be transcribed can be converted to text easily, all you have to do is proof read and format the output. In our tests a few years back we calculated a time saving of some 60+%
So lets see a couple of examples. (tip, play in HD and play in the YouTube viewer):
Example 1. Voice-to-text single speaker audio. This audio was taken from an ABC Australia TV show so obviously no ability to train the software to the speakers voice. The speaker is Leigh Sales:
Example 2. Voice-to-text one on one meeting. This is a brief chat between ABC Australia journalist Leigh Sales and leader of the opposition Bill Shorten.
- Audio ripped using Audio Hijack
- Videos created using Camtasia for Mac
- Voice recognition software is Dragon Professional for Mac v6 (SKU: S601A-G00-6.0)