Some languages have more regular prosidy than others, or are more phonetic than others.
English and French have a common phonomenon that doesn’t really exist in Romanian (and probably not much in German, Dutch, Estonian or Hungarian either.) That is homographs. Two words, with different meanings, that spell the same but have different meanings. Sometimes, the AI can guess the pronunciation of a homonymn by its context, but often AI gets it wrong. One example in English is read. If pronounced one way, read is in the present tense, another way it happened in the past.
I read every day.
Alone we would assume it is present tense, and would sound like reed. But in a paragraph, read would be in the past tense, and sound like red.
When I was a teenager, I loved books. I read every day. Now, I don’t feel like I have as much time, but somehow I still end up watching a lot of series.
In English, the prosody can change the meaning of a sentence. The prosody and pronunciation of homographs are often obvious to a native speaking humans. But, what about when a computer tries to tell us how to pronounce things? One confusing article about how to pronounce Marriot shows the AI making all kinds of mistakes.
(In short, Marriot rhymes with chariot.)
But, listening to the AI, we hear humorous inconsistencies. It is easier