Speech synthesis based on a neural network?

Question

Neural networks

Hello. Interested in the topic of speech synthesis from text using neural networks. The voiced language is Uzbek - no one was engaged in its synthesis. For a couple of days now I have been reading articles on the topic of TTS, smoking docks))
I stumbled upon this resource . I think to do what is written there, but I feel that everything is very superficial. Is there anyone who is closely involved in text-to-speech? Am I choosing the right machine learning Torch tool? Recommend articles for basic speech synthesis of the "new" language.

Answer 1

The first torch (which is in lua) is already history, look at pytorch. But in general, in the comment above, they correctly centered that a lot depends on datasets, and there are most likely no public datasets of Uzbek speech.