I had a great idea that was terrible

I had a great idea a couple of months back. Really great. So good and easy to execute in fact, that I immediatly put it on top of my “side-projects”-list.

So let me pitch you the idea: A podcast for people who like to get a quick daily update on what’s happening, fully automated. The idea was simple:

  1. Get the three most popular news articles of the day/evening via some API that’s surely out there
  2. Isolate the actual text from the HTML.
  3. Let a text-to-speech engine read the texts and safe the result as a file.
  4. Maybe enrich the file with a jingle or whatever.
  5. Upload the file to a podcast hosting service.
  6. WIN!

Set it up once and just have it running automatically for years. Sounds awesome, right?

Some of the people I was telling about this idea had concerns about the quality of text-to-speech engines and if you would actually want listen to them for a period of three full news articles, but I was thinking about all the voice assistants out there (Alexa, Siri et al.) who will already read you your news, if you ask them to, so it wasn’t really a concern for me. So one lonely afternoon I started to work on the project, step by step.

  1. I skipped this part at first since I thought it would be one of
    the easier parts. I just grabbed a url of a random news article and went
    with that.
  2. I used txtify to get that done, piece of cake.
  3. There is actually a python package, that uses Google’s text-to-speech engine and offers all the functionality you would want.

Using this package to produce an mp3 file I listened to what I had just created, at this point I was about two hours into the project. The result was, well, underwhelming, to put it nicely. Not only was I sick of the voice about 20 seconds in, I also could not for the life of me recall the contents of the article. It was so bad in fact, that I stopped the experiment there an then. I wish I could present you a big universal learning from this, but there really isn’t one, except maybe that modern text-to-speech software has impoved surprisingly little ever since I played around with Microsoft Sam on Windows 95.