> For example games have exactly zero accessibility support, system actually knowing about what is being rendered is quite useful for it to provide tools like voice over.

Some engines do have accessibility support. Most games don't do accessibility because it costs money. Same as web apps, by the way. It is not a technical limitation.

> What about IME?

What about it? It works just fine in any proper engine. There are a lot of non-Latin-charset speakers in the world...