I just listened to the slides set to use a Japanese speaker. (The file was shared with me because I've worked with Jackie and her team).
For the first slide, the voice speaks the English text that is in the Text-to-Speech object. However, because the program is set for Japanese, the English words are mispronounced, in addition to being spoken with a Japanese accent. So the English text sounds foreign, but it's actually still in English. (The numbers are spoken in Japanese, because they were entered as numbers.)
Bottom line: Text-to-speech does not translate text. Changing the language in the drop-down merely tells the program what language to expect. The program then pronounces the text based on the pronunciations typically used in that language.