I listened to the whole song and about half the words sounded synthesized to me, while the other half often sounded like they started with a real human voice and maybe were subject to autotuning or heavy postprocessing.
Still, there were at least portions of the song where I would have been fooled and said it was a human being singing!
(But maybe the prevalence of autotune is actually gradually lowering my standards for what sounds like human signing...)
On further listening I think there are some specific syllable codas that sound extremely synthetic, while there are others that sound tolerable. /s/ as the coda seems pretty uniformly bad, while /n/ seems quite good.
More listening: the r-colored vowels are pretty uniformly crummy; open syllables are good. I found the words "reflection" and "connection" (in the song "Run Away") pretty impressive. Suprasegmentals are often a weak spot.
Still, there were at least portions of the song where I would have been fooled and said it was a human being singing!
(But maybe the prevalence of autotune is actually gradually lowering my standards for what sounds like human signing...)
On further listening I think there are some specific syllable codas that sound extremely synthetic, while there are others that sound tolerable. /s/ as the coda seems pretty uniformly bad, while /n/ seems quite good.
More listening: the r-colored vowels are pretty uniformly crummy; open syllables are good. I found the words "reflection" and "connection" (in the song "Run Away") pretty impressive. Suprasegmentals are often a weak spot.