The Text-to-speech always select the acoustic rendering of a word that is most likely to be correct, given a particular sentence. However we might in some case want to get an alternative rendering of a word, maybe to give more or less prominence to a word, or to correct some acoustic issue.
Now it is possible to get alternative rendering of word simply by typing @alt1, @alt2 or @alt3 before the word in question. Like this example:
What a @alt1 strange world.
In this case we are intructing the text-to-speech to use an alternative rendering for the word 'strange'
The effects of the alternative rendering are often subtle and are difficult to predict in advance, so you need to use a trial and error method.
Watch this tutorial video for an example of how to use alternative renderings of words: