The AWS example is much better, hands down. All the voices in Unreal have lots of artifacts, and the diction is subtly but noticeably stilted.