BTW. I can't optimise any languages except English, so its likely that all non-english languages may not announce things as well as the English version does
If there is a problem with any non-english language version, someone who speaks the language will need to use VoicePolly.pro to test some alternative spellings of words etc to make the syntheseised speech sound OK.
e.g. In the English version the word "Watt" was being synthesised as "Waaaa", so I changed its spelling to "Wattt" and this seems to have fixed the problem in English, but I have not changed any other languages.
I also had to use a "Neural" voice for some letters of some voices, and manually replace the MP3 files, but unfortunately "Neural" voices are only available for English UK and USA, and Spanish USA and Portugeuse Brazilian.
And Neural voices are not available from TTSMP3 or VoicePolly, I had to log into AWS to manually create the words or letters which were not correct.