Tidbits on software development, technology, and other geeky stuff

Text to Speech through Airtunes

A few months back, I setup a small Linux server (PogoPlug pink version running PlugApps Linux) which has been up and running in my front closet. The thing is silent and uses only 4 watts of power. It’s one of my favorite toys. I plan on writing a post later with details on this bad boy, but for now I mention this as background for this post.

I’ve recently wanted to figure out a way to get this server to send sound to my Airport Express (one of Apple’s better products I must say) so I could effectively setup an “alarm clock” that would play music from my music library. You know, a cron job every morning at 6AM that blares my favorite music at me so I will get out of bed.

It took me awhile to find it but I finally came across JustePort which is a .NET app that streams music to an Airport Express. The author, Jon Lech Johansen, did a great job with this and kindly provides the C# source code which is very interesting to go through. Using Mono, I can easily use this .NET app from within a Linux environment.

Anyway, setting up a cron job to call JustePort and pass in some music files was pretty easy to get up and running. In fact, it works great. But I wanted to take it one step further. I wanted my Linux server to be able to “speak” to me through the Airport Express. This enables some interesting scenarios. I could get it to say “Wake up sleepy head!”… followed by some music. Or, it could poll my email account and say “You’ve got mail”. Or, I could setup recurring reminders so that once a month it says “Change the water filter!” from the speakers setup in my living room. Necessary? Not really. Geeky Cool? Absolutely.

I finally settled on using The Unofficial Google Text-To-Speech API, which responds with mp3 speech given some text, to generate the audio. I looked at using espeak or festival but the natural voice quality of the Google service is much better. Once I figured out how to use wget to download a mp3 of some text, I needed to pad it with some silence on the end because I found that if the audio was less than 8 seconds (or so), JustePort (or the Airtunes) wouldn’t play it. In addition to padding, I found that I needed to change the sample rate because when I played it, it sounded like chipmunks who inhaled a couple of helium balloons. The awesome sox utility came to the rescue to perform the padding and sample rate conversion.

With the help of Google Translate, wget, sox, Mono and JustePort I can send dynamic text speech to my Airtunes device and effectively enable my server talk to me.

Here’s the single command (heavy piping!) to say “Hello World” through the speakers connected to my Airtunes device.

wget -O- --user-agent="Mozilla/4.0"
  http://translate.google.com/translate_tts?q=Hello+World | sox
  -t mp3 - -t raw -r 87k - pad 0 4 | mono /opt/JustePort/JustePort.exe
  - -15

A few notes:

Discuss on Twitter