Monday, November 03, 2008

Enter the Dragon

We read a review today on Dragon NaturallySpeaking 10. For those who have never heard this software, it is one of the leading voice recognition and translation applications. We have long wished that we could just speak to our laptops and be able to instantly capture our thoughts. Alas it takes time to train the application and the task of turning talk into text is often slow and not error free.

My blind mother in law has used the Dragon software with ease having written around half of her 30 plus novels using its earlier versions. When you witness an expert using it one can easy see the benefits it offers. She also uses the voice recognition on her mobile and never needs to touch the keyboard. It is interesting that every mobile has this feature but few take the time to use it.

Text to voice is the reverse process and also has been around for a number of years. Today’s applications in this area are very sophisticated and are built around huge voice databases and the ability to style-sheet voice to text. A couple of years ago we looked at this software with a view to being able to produce audio books from digital text. The template building was somewhat of an art form and involved specifying the pitch, time delays etc when certain punctuation or narrative was detected. Once style-sheeted the text was processed against a voice engine and a file created. This file could then be refined by a sound engineer and the total process completed in a fraction of the time and cost of traditional production. The results were not ‘robotic’, but clean and crisp. They often dealt with complex scientific terms with ease, but often ran out of puff on the long or badly punctuated sentence.

We didn’t take it further at that stage but constantly monitor the progress of the leaders such as Loquendo in this area. We suggest you go to their site and try it out for yourselves.

No comments: