Tuesday, October 26, 2004

Creating an audio blog reply using Festival

I listened to Dave Winer's Coffee Notes for 24-Oct-2004 podcast on the way home tonight and felt compelled to reply to him with an audio reply email about Richard Scoble/Scooble's voice mail. Here is how I created it.

* Get the Festival Speech Synthesis package for your Debian GNU/Linux box


$ sudo apt-get install festival

[...bunch of status about installing a festival and dependancies removed...]
$


* Create a snappy reply. I like emacs. This is exactly what hi-dave.txt looks like:

Hi Dave!

This is Marc No-zell and I just listened to your Coffee Talk with
Richard Scoble. It sounds like Microsoft's voice mail system is using
a descendant of the DECK-talk system that was created by DECK in the
1980s.

I'm creating this audio reply using the Festival Speech Synthesis
System on my Deb-ian GNU Linux laptop. Neener, neener, neener!

Well, thanks for the Coffee Notes!

Now I did do a little cheating here. Festival follows strict pronunciation rules and sometimes gets words wrong. For example my surname would be said as 'nozzel', so I spell it how I want it to sound. The same is true for DECtalk and DEC, which sounds like 'dec-tork' and 'december' (trying to be a little too clever ;-)

* Convert from text to wave format:


$ text2wave hi-dave.txt -o hi-dave.wav


* Convert from wav to mp3 format


$ lame hi-dave.wav hi-dave.mp3



* Listen to the result:


$ xmms hi-dave.mp3