I’m having an issue with the advanced/AI voices (that come with premium plus) in Arabic. For years written in digits (which regularly use Arabic numerals, confusingly the name for the ones in the English alphabet), the text to speech renders incorrectly. Sometimes it is in accented English, sometimes it is total gibberish (to me, but I suspect it might be Farsi).
Appreciate if this could be looked into - I know there is auto-detect language for ElevenLabs voices, perhaps it is doing this instead of being set on the specific language? This has happened across the last few articles I have imported/speech generated.
Hello @Trafs Could you please check exact voice in the Settings/Reader/Text-to-Speech? And also do you generate audio for all lesson at once? Or do you use Sentence view? I would appreciate if you can add link to the lesson too. Thank you in advance!
I’ve changed to an alternate voice which seems better but still has issues. An original article here it defaults to either gibberish or English is here: Login - LingQ , whereas I’ve tested and it seems to handle Hindi numerals (i.e. written in arabic script, ١، ٢، ٣) much better - Login - LingQ than those written in Arabic numerals ( confusingly the name given to these: 1, 2, 3) Login - LingQ .
I’ll stick to Hindi numerals in the future regardless where possible, but a lot of news websites and other sources for imports use arabic numerals so imagine it’ll continue to be an issue.