Closed mircealungu closed 6 months ago
Indeed, we need to specify the voice we like for every one of our supported languages!
However, there's now a fallback: if there's no preferred language that we specify, then we fallback on the Standard-A which seems to exist for most of the languages: https://github.com/zeeguu/api/blob/36f663e3095946f44cc2e3ab7e6a954bdb389636/zeeguu/api/endpoints/speech.py#L25. Although it sounds quite bad so we should select defaults for all the languages we care about.
Which brings me to, we still need a voice for Portuguese: do you want to choose one that you like? You can choose from here: https://cloud.google.com/text-to-speech/docs/voices
The language to voices mapping is here: https://github.com/zeeguu/api/blob/36f663e3095946f44cc2e3ab7e6a954bdb389636/zeeguu/api/endpoints/speech.py#L12
That's a good strategy, I had a look at the list yesterday and it seems there are various models to choose from.
I think regarding Portuguese, pt-PT-Wavenet-A
sounds good. I have heard that voice a lot in applications, so I know where it comes from!
I think a good feature in the future would be to add a "Speak Slowly", which we could do by playing back at 0.75.
Speak slowly feature is a good idea. I'll merge for now this version of the web, and we can still add and change voices on the backend later.
asking for your review @tfnribeiro