Why does speech synthesis work well only over the Internet?

Question

speech synthesis

Faced with the fact that a more or less high-quality reading of the text is obtained only from Yandex and Amazon and only via the Internet on their servers.

The quality of synthetic speech cannot be tolerably organized without an Internet connection?
How it works?
Or is it just more convenient to connect to an Internet to sell?

Answer 1

They keep their secrets. That's all.

Answer 2

Of the available / open / free Russian language, the rhvoice engine synthesizes with high quality .
Windows 8 and above perfectly synthesize offline English, Russian is also not bad, but it's very depressing, it's so hard to listen to.
ps The arrangement of stresses and accents in the text helps to synthesize speech qualitatively, i.e. for speech to be expressive, you need to highlight it with a lot of intonation, there are a lot of tools for this in synthesis engines. The trouble is that for this to happen automatically, you need to create such a good AI, and of course, companies such as google / yandex / microsoft will not share such competitive information.