On Thursday, Microsoft researchers announced a new text-to-speech AI model called ... and content information respectively. Finally, the generated acoustic tokens are used to synthesize the ...