Did you know Japan had singing computers? I sure didn’t, at least until I stumbled on a report from this morning indicating that they not only did have machines that’d outperform half of American Idol’s starry-eyed hopefuls, but that a team of scientists from Tokyo have improved the program’s artificial algorithms to make the machines sound more human-like.
Here’s what one of the programs, VocaListener, currently sounds like:
Sounds pretty close to human, right? I sure wouldn’t be able to tell it was a robot on first listen. (Plus it’s catchy in a J-pop/guilty pleasure sort of way.)
But according to Science Daily, a few of the researchers from the Graduate School of Engineering at The University of Tokyo have found a way to make the inflections sound more human. The vocals posted above are built from a single type of frequency curve, meaning there’s little variation when you go from note to note. But the graduate researchers have programmed in a new type of frequency wave that has a lot of the same tics as human singing. Essentially what they’ve done is recreated the nuances that lend our voices their character.
Kind of weird, especially when you consider most musicians nowadays use auto-tune to sound more like machines.
(via Science Daily)
More on TIME.com: