Cepstral Community

We use your voices with a program called automate (www.networkautomation.com) to allow their program to talk and let us know what automated function is going to run and when it is running, this has resulted in a very reliable way of doing repetive tasks that automation can handle while still making sure everything is running fine by being talked to while the task is running.



The combination of the 2 programs results in a fully automated speaking computer assistant....

Instead of some simple tone, I have configured some notifier tools to call upon a voice snip that intelligently informs me what kind of new item to check, when I am across the room deep into a book.

I use MILLE, AMY LAWRENCE AND WILLIAM as Pilot and Controller voices in the ATC program VOXATC, this provide as close as a ture to life atc interface as you can have ina flightsim without going live with VATSIM. I am a former pilot, so I apperciate something better than the FSX default interface

I use my Cepstral voice in my car. I have a Pontiac Firebird converted into the likeness of KITT from Knight Rider. I had been using William, but just got Lawrence for a change of pace.
The text-to-speech gives me more flexibility with the program I'm using instead of just using .wav files from the show.

yep its wonderful program.

I use Cepstral for shorthand practice. Typing a passage and using TTS is much easier than recording a passage a at target speed and using a sound editor to get speeds on either side. We're about half-way through automating it.

The group I'm in needs speeds from 20wpm (beginners) to 200wpm. We usually bracket our target speed with two speeds on either side (usually 10 and 20wpm).

My husband wrote a program that adds <break= > tags to a text file between each word. That worked well enough for my personal use, but shorthand writers usually think in terms of wpm.

We're now experimenting with break times and voice rates to get specific speeds in wpm. Simply setting a low rate results in words that are dragged out to incomprehensibility. Simply setting the break time results in silence with staccato words.

The end goal is an program that, when given a passage and target speeds, will create .txt (or .ssml) files and a .bat to send them all to Swift to create a sound file for each speed.

Unless there is huge demand (unlikely), it will be up to the end user to license their own voices. They will also have to provide their own text files, as most shorthand books are still under copyright. Still, licensing their own voices and typing the passages is inexpensive and easy compared to the other ways we've tried.

Many thanks to the Cepstral support person several years ago who suggested we check the SSML page on Cepstral's site. None of the other TTS companies bothered to respond.