Different types of Portuguese

[borisservo]

Hello,

I'm struggling here a bit, the software works fine and I've been using in English.
When changing to Portuguese that another story.
In Portuguese we have different accents and dialects for different countries.
Portuguese from Portugal
Portuguese from Brazil
Portuguese from Angola
Portuguese from Mozambique
and there few more.

The one that is imported to the this project is from Brazil that drives me nuts because the wording is off and pronunciation is even worse.

Is there a way to import different types of Portuguese languages just like in the Ubuntu or Windows deployment is done.

Thank you.

[ROBERT-MCDOWELL]

it has nothing to do with E2A, it concerns the models used buy each TTS engine.
If you want a portugese model (which TTS again? E2A manages for now 6 TTS engines!)
so find one and we will put it on the list as fine tuned model.

[borisservo]

Hello @ROBERT-MCDOWELL,

I'm aware that it has nothing to do with E2A, for that reason I was asking if there was a way to import.
Changing the code somewhere or using the dataset collection link for example.

This is TTS that uses Portuguese from Portugal.
https://hf.co/collections/Jarbas/onnx-tts-community-models

Thanks in advance

[ROBERT-MCDOWELL]

you must find a way to convert this onnx model to xttsv2 if it's what you are looking for.
or create your own xttsv2 model using our training model.

[ROBERT-MCDOWELL]

btw I can maybe wrap onnx models and use them with VITS engine. but for that I need you to test.