Sometimes when I add Text to Speech audio to one of my lessons, it really mangles it. I especially get frustrated when I am doing a fill-in-the-blank matching board--it reads the "____" blank I type in as "Underscore, underscore, underscore.." Can I do anything to fix it?
Mispronounced in Misenheimer
First off, let me commend you for playing back your recording to check it! All too often, I come across lessons submitted to the library where the creator obviously hit that button and was done with it (unless of course they meant for it to sound like a robot having a psychotic break with reality...) And let me assure you that yes, there are ways around the occasional verbal burps the Viz sometimes belches forth with her Text to Speech recording.
Sometimes you have to enter the text phonetically, record it, and then go back an fix the text to make the TTS behave. This is especially true for an unusual name, or if Viz has chosen the wrong pronunciation of a homograph (like you meant the pretty ribbon bow, but Viz put it at the front of a boat...).
And, as you found, fill-in-the-blank questions can be problematic as well. For those, I usually type in something like, "The boy is ... blank..." and record it. (The ellipsis gives it just the right pause to set off the blank.) I then record it with Text to Speech (playing it back to double check it, of course!). Once it is right, I fix the text, changing it back to "The boy is _______."
I babble on about it all ceaselessly in the video Audio: Importing and Creating Audio Clips if it would help you to see it done.
Hope this helps!