Originally published at: http://boingboing.net/2017/04/29/synthesized-singing-voices.html
…
I dunno… the male voice was pretty obvious. The others were pretty good though.
Well good. It’s high time singers went the way of coal miners and farmers! Maybe they can be retrained as… I dunno, Soylent Green?
They sound like English vocaloids with not very good tuning.
They are like people with naturally good singing voices who have never heard singing before and learned how to sing from reading a book.
That’s what uncanny valley sounds like.
HEY I’VE HEARD ABOUT THIS PROJECT! This is a singing synth engine being developed by the folks over at Voctro Labs and they have a bit of experience making spanish Vocaloid voicebanks. In fact the female spanish voice on that page is the same one used in their Maika bank.
By the way you can read the more technical details about it over on it’s research paper https://arxiv.org/abs/1704.03809
Yeah, the male voice is weirdly bad, given how much better the female Spanish voice is. I’m not sure what’s going on there, especially since, if I’m reading their paper correctly, they used quite a bit more data to train the English voices.
If I had musical talent I’d remix the acapella voices into vaporwave.
I’m pretty sure you could have just about anyone to stand up there at 1AM, lip sync to ‘Take the A Train’, and get a bunch of boozed-up, wizzled lounge lizards to happily applaud.
Politicians - the last refuge of any scoundrel.
I’m a singer. I sing to people in assisted-living/memory-care centers.
That’s important. I did that myself, long ago.
This topic was automatically closed after 5 days. New replies are no longer allowed.