I just followed a link from a plone product (Speechcast 0.01) to the webpage of the MARY text to speech project at http://mary.dfki.de/. This is a very promising project that aims to build a open source text-to-speech system, based on a JAVA server client system with the ability to build webbased services. I was completely surprised how well this works. The first few voices i tried were just like the normal “robots”.
But if you like to hear how text-to-speech works nowadays then head over to http://mary.dfki.de/online-demos/java-interface (you need java 1.5 web plugin) and select one of these voices:
- time-awb (English, male, limited)
- dfki-zeit (German, female, limited)
- dfki-bundesliga-excited (German, male, limited)
- dfki-bundesliga (German, male, limited)
i never heard such emotional computer generated voices – that’s the right way! Imagine what this could do for blind people or how you could generate automatically podcasts from your texts or how you convert a book for studying on your ipod 