Podblogging, or using TTS to create a podcast

I recently read an article by a Darren Barefoot complaining about the quality of the material on podcasts compared to blogs. And just today I was thinking how nice it would be to have a podcast of daily biographies of great people, and how this sort of material is probably available on the web in written form only.

Well perhaps text-to-speech (TTS) can help us to bridge the gap between the high quality of textual material on the web and the convenience of listening to audio on a portable device. What if we used TTS to convert the webpages we want to track into audio files we can listen to during the commute? Input: text, output: audio. Shall we call it "podblogging"?

Update: I have cobbled together a primitive podblogging system, and the results are fantastic! Basically, whenever I see a webpage that I want to review later, I tag it using del.icio.us. My podblogging system then converts it into an MP3 file and downloads it to my PDA. I've uploaded a sample to OurMedia that will hopefully be processed soon so you can listen to the results (the page being read is http://linuxjournal.com/article/8217). Effective immediately I am unsubscribing to all my podcasts and listening exclusively to podblogs!

Update 2: I have released an open-source Windows program called Audiolicious that converts any RSS feed into a podcast, using text-to-speech. Hopefully this will help to bring podblogging (or "textcasting") to the masses.

This proof-of-concept system that I have assembled is a pretty complicated bit of geekery, but it's still cool to know that it can be done. I'm using a bunch of free programs:

• lynx -dump to convert from HTML to plain text
• pyTTS to convert from text to wav
• Lame to convert from wav to mp3
• MobSync to move the mp3s onto my Pocket PC

If anyone wants a hint on making their own podblogging system, here are the scripts I use. Not too polished or portable I'm afraid, but it might give you some ideas that you can take away.

----- delicious-audio.bat -----
ruby delicious-audio.rb



----- delicious-audio.rb -----
class Array
  def my_shuffle!
    size .downto 1 do |n| push delete_at rand(n) end
    self
  end
end
Dir['c:/Documents and Settings/Jon/My Documents/X30 Storage Card/Delicious/Audio/*.mp3'].each { |file|
  File.delete(file)
}
Dir['c:/Documents and Settings/Jon/My Documents/X30 Storage Card/Delicious/*.html'].my_shuffle![0..4].each { |file|
  file.gsub!("/", "\\")
  system("copy \"#{file}\" delicious-audio-1.html")
  system("lynx -dump -nolist delicious-audio-1.html > delicious-audio-2.html")
  system("txt2mp3.bat delicious-audio-2.html \"#{file.gsub('.html', '.mp3').gsub('\\Delicious\\', '\\Delicious\\Audio\\')}\"")
}



----- txt2mp3.bat -----
@c:\python23\python txt2mp3.py %1 %2



----- txt2mp3.py -----
import pyTTS, re, os, sys
input_file = sys.argv[1]
output_file = sys.argv[2]
tts = pyTTS.Create()
tts.SetVoiceByName('MSMary')
file = open(input_file, 'r')
text = file.read()
file.close()
text = re.sub(r"[\n\t]", " ", text)
text = re.sub(r"\[[^\]]+]", " ", text)
text = re.sub(r" +", " ", text)
sentences = text.split(". ")
i = 0
# Set output_file file to 0-length [Jon Aquino 2005-04-03]
file = open(output_file, "w")
file.close()
for sentence in sentences:
    sentence = sentence.strip()+"."
    i = i + 1
    print str(i)+"/"+str(len(sentences))+": "+sentence
    tts.SpeakToWave('tts.wav', sentence)
    os.system("lame --quiet -b 32 tts.wav tts.mp3")
    os.system('copy /B "'+output_file+'"+tts.mp3 "'+output_file+'"')