the TTS assumes english when the word exists in english.
But you know very well that some english words derive from other languages, such as French.
As such, how do you assume the word is english, when it didn’t even come from the english language in the first place.
That is, the english assumption is a flaw in any case where we do not use english TTS
Here’s why:
https://platform.openai.com/docs/guides/text-to-speech/voice-options
“The current voices are optimized for English.”
Your post is like if I say “the problem with the TTS optimized for Japanese is that some Japanese words come from English. So when I try to get it to speak English it is not good.”
Fixed:Here’s why:
We cannot specify a language in the query
Rather, the model and tuning is not trained by polyglots that read 100 audio books in every world language.