Just a suggestion. If you could have ‘voice/language pairing’ as an option. To be honest no matter which languages you choose the main voice will not sound good speaking it’s non-native language. IF you could pair an English voice with English, a Hindi voice with Hindi, a German voice with German, etc. The result would be superior. The fact that it’s a different sounding voice is minor compared to having an English voice try to speak Cantonese and sounding like a Duo Lingo dropout.
Hello @Bob_Hawkey,
Thanks for the suggestion! To better understand, I’d like to clarify—currently, voices are listed based on the selected language. If you choose multiple languages, only multilingual voices are displayed.
Which specific voices are you experiencing issues with?
Thanks for responding! I fear there are too many language and voice combinations for me to check them all out. Plus my knowledge of languages other than English is very bad. However, I have run into several cases where you could select several languages - English, Cantonese, Italian, German as an example. Then you would select a ‘multilingual’ voice that you like but when you tell the character to speak to you in Cantonese or Japanese it is exactly like an English person reading some Japanese words rather than a native Japanese speaker talking.
I can imagine that this type of thing is really, really difficult to implement and I don’t envy you in the least having to do it. It absolutely does happen though. The non-english corollary is when you might select a multi-lingual German or Italian voice and it tries to speak english. It will do so but sound like a German speaking bad english. I actually like that but a lot of people do not.
To my mind the only solution would be to support some sort of voice switching where a German voice would speak German, a Japanese voice would speak Japanese…etc, etc. I know this would add complexity to the voice support system. You would have to filter voices by language and then make sure the correct voice/language pair is chosen at run time. This would mean essentially a different voice for each language but at least the inflections and accents would be much more realistic.
You guys are doing a great job and I want you to succeed. I’m not sure this methodology would work or improve things but it’s something to consider perhaps. Keep up the great work!