Computer downloads random audio streams from internet. It then applies a speech recognition engine to transcribe those streams into English text - regardless of source language / type.
Concurrently it constructs a voice by cutting up the audio streams themselves into individual phonetic sounds which it sorts into folders.
It then speaks it's English text by pasting together sound files randomly drawn from the appropriate folders. As soon as a phonetic file is used it is erased.
The whole process is continuous and endless.