Okay, this kind of like watching “America’s Most One Sided Fist Fights”, but I was diddling around with Cspeak, which is a speech synthesizer that is just 150 lines of code long, and thought I’d try to see just what quality it was capable of.
Well, it’s (not surprisingly) pretty hard to understand most of the time. Try this out:
Pretty hard to grok. It’s not just that the consonants are poorly represented, it’s got all sorts of oddly confusing changes in prosody and pitch which make the sentences really difficult to track. Did you figure out what work of fiction this was from?
You could cheat by listening to the same synthesized from Festival.
It’s interesting to surf over to Dennis Klatt’s History of Speech Synthesis and listen to some of the historic speech synthesizers, and decide where CSpeak falls in terms of legibility when compared to other historic speech synthesizers. I also liked this page on pre-electronic and pre-digital attempts at speech synthesis, which had some remarkably cool machines dating back to the 18th century.