Cepstral Swift Voices Version 4.0.1 - Text-to-Speech for Mac OS X

reviewed by Robert Pritchett

We reviewed Cepstral voices back in April 2005; http://www.cepstral.com/cgi-bin/news?page=2005-03-18-00.  Since then, Cepstral has not stood still. They have continued to advance forward, expanding out to more voices, improved the multi-lingual cross-platform rendering engine and made voices available for mobile equipment. Pronunciation and text processing have improved as well. Even 64-bit support has been added and the voices work with the Mactels, Sparc and Solaris, as well as the Windows and Windows Mobile systems.

Most Mac-apps co-habitate with Cepstral voices. If you find one that doesn’t play nice, they want to know about it. And if you don’t like the pronunciation on the Mac, here is a link to what you can do to “make it so”: http://www.cepstral.com/cgi-bin/support?page=lexicon. There is a lexicn.txt file that can be manipulated. If you really want to delve into homonyms, homographs and phonemes, be my guest. Cepstral made the science of digital voice easy.

Boy, that’s real Swift! If you want to get geeky, go visit the Support area on the Cesptral website and get into the FAQ sections depending on which operating system you use.

My hat’s off to Dr. Alan W Black, Kevin A. Lenzo and their voice-enabling magicians to make this work as well as it does across platforms and within various technologies and human languages.

If you want to keep up with all that they have been doing, go here:

http://www.cepstral.com/cgi-bin/company?page=news.   

Each voice can be downloaded and a 7-page QuickStart Guide is included in each .dmg for the Mac. I can only assume the package is similar for other Operating Systems.

The version 4.0.1 voices I was able to download for Mac OS X (and it took a while) were:

I didn’t pick up Dog, Shouty or Whispery.

The download files for Windows is smaller per voice and for Linux a little smaller still. There are also separate voices for Windows CE systems.

My personal favorites are David for a male voice and Diane is still my favorite female voice. Both are “close enough“ to natural that I think anyone with speech challenges would be happy using them in assistive applications or augmentative and alternative communications (AAC) apps.

Oh, come on! You always wanted to make an app speech-enabled and take the existing Mac OS X speech functions up a notch. Now you can!

And of course, Cepstral doesn’t do this just for the Mac, but also for navigation aids, mobile communications, toys, games, medical instruments, weather systems interaction, industrial apps, and government even has a voice. They are stretching themselves into all kinds of text-to-speech and speech-to-speech apps. Think Star-Trek-like universal translator kinds of things. It won’t take too long for them to get there. Some military apps for translators are already being used.

Each voice has a $30 USD license, but you can try before you buy. And these link we researched before are still worth repeating:

Dig Deeper

Cepstrum - http://cnx.rice.edu/content/m12469/latest/

http://www.cepstrum.co.jp/1about.htm 

Cepstrum Analysis -

http://documents.wolfram.com/applications/signals/CepstralAnalysis.html

http://tiger.la.asu.edu/software/cepstral.htm

http://htk.eng.cam.ac.uk/docs/cuhtk.shtml

http://svr-www.eng.cam.ac.uk/~ajr/SpeechAnalysis/node33.html