Text-to-Speech with Cepstral

I receive a daily opinion e-mail that isn’t critical for me to read but I enjoy it and like to get to it when I can. It’s about a 15 – 20 minute read but my schedule doesn’t allow that time slice every day. However, every day I have at least 15 – 20 minutes where I’ll be commuting, doing chores, or the like, and that is when I’d like to catch up. For several years I’ve been using Text-to-Speech (TTS) software to convert the email to a “spoken” sound file. At first I was using Microsoft-provided voices (years ago) which, then, had a distinctly robotic characteristic. The technology has come a long way since then.

For the software, I started using Nextup‘s TextAloud software, which worked well. However, as my platform requirements changed, I found myself back on Mac as my main platform and needed a Mac solution. I found the company Cepstral, and their voices sound very good. While still not sounding like human speech, the technology has progressed much. Cepstral also supports Mac, Windows and Linux, which gives me better platform independence.

Cepstral supports Speech Synthesis Markup Language, or SSML, which “give authors of synthesizable content a standard way to control aspects of speech output such as pronunciation, volume, pitch, rate, etc.” My needs are simple, I use it to add a pause after a paragraph. This is helpful so that I can tell where paragraphs end. So I have to prepare the text for reading. Here’s what I do.

1. Copy the text from my email client into BBEdit
2. Run a Text Factory I’ve written on that text that prepares it for TTS.
3. Save that file to my desktop
4. Run that file through the Cepstral command line language application “Say.” That application will produce a corresponding AIF sound file of the spoken text.
5. Open that file in iTunes, which copies the file into the correct Smart Playlist

This is the MP3 file of this posting, using the above process: 081014_blog-posting

Here are the two supporting files for this process:

say_this.sh:

#!/bin/zsh
say -v "Cepstral Diane" -f $1:r.txt -o $1:r.aiff
open -a iTunes $1:r.aiff

My text factory file: