Continuing my quest to fill up the unused parts of my brain with useless crap, I stumbled across IBM’s ViaVoice TTS system for Linux. It is used to convert text files in to a synthesized voice (which is part of what Lenzo is doing with Sphinx). Out of the box, it works ok. My problems started when I wanted to create mp3 from this system. When some recompiling, it is possible to get one of the command line programs to create AU files, rather than stream noise to /dev/dsp. Unfortunately (I blame pthreads here), I find that not all of a given input file gets successfully converted. If this were in Perl, I’d simply write $|++ somewhere. But, this is C++ and there are libraries for which I do not have the source code, so I’ll just pound sand. If anyone can suggest a workaround, or another system to convert text into speech which runs on the command line and produces AU or WAV or even MP3 files, I’d be much obliged.

Of course, the goal here is to create RadioFreejjohn, in which my Apache::MP3 server also reads me RSS feeds, thereby similuating a real radio station.

As I said, a good time killer.

[Original use.perl.org post and comments.]