Some text are read better by Simon than by Dave. Some text are instantaneously read very naturally without having to using the time consuming tags.
Though time consuming, Through the use of tags inside Loquendo TTS Director it's possible to read any text as natural as a human.and "accent less" (Received Pronunciation).The pros of using TTS vs Voice Narration: AT&T Natural Voice(s) would be #2 and Acapela Telecom voices #3 The voices available from Loquendo (version 7+) are the #1 I have seen/used. You can make your own dictionary for words you would like to have pronounced in a customized way, it's also possible to have words spelled, or say number in time/date formats. Via speech tags, you can manipulate speed, pitch, duration, insert various pauses etc.
My first conclusion would be that the best software (writer) to be used with voices is "Loquendo TTS Director", this software will give you complete control on the voices. Having tried various tools, including the aforementioned textaloud. My experience with TTS: (not a scientific backing the learning lessons are for commercial use I'm (unfortunately ) Italian so I need tts with italian voices Nevertheless I think that it's better than nothing and it could be a useful service for users with problems like dyslexia. But it's not worth the hassle: it's less expensive record real voices). That said, I know that it's almost impossible to realize tts audio files that seems real human voices (I already worked with Loquendo, an Italian company acquired last year by Nounce, and Acapela: with a very hard work you can reach a very similar result. Only if it's impossible, I would consider the solution of exporting the texts, using a tts to produce an mp3 file of each text and importing the audio files in Storyline. I try to fix: I would like to insert a button inside each page of the Storyline lesson to make a text to speech on the fly when the user click on the button. I realised that I didn't explain in a very clear manner what I need. Thank you all for the very interesting answers.