I'm familiar with most of the material in this video but not that Chinese singer--I don't know what the real singer sounded like but the replication was free of any robotic nuance and gave me goosebumps.

I love the video's narrator Elizabeth Zharoff's cosmic "yes" about all the possibilities of AI that sings. Just the right amount of concern for the legal implications as well.

I messed around with Holly Herndon's "Holly+" online software last year. Found it hard to get really clean results, but I honestly didn't try too hard:

https://holly.plus/

Just upload an audio file of an a cappella of your voice and it will be replaced by hers.

I love that she mentions Yamaha Vocaloid and singing synths like the popular anime recording artist mentiioned. Vocaloid has been around for over a decade and it's a software stepping stone to the current state of singing AI.

I have used many VST virtual (sample based) instrument/singers over the years. The problem, other than mostly being vowel sounds, has been getting convincing "legato" singing, where the transitions from note-to-note sound realistic. Only the more recent ones seem to have figured out how to do this well.

No mention of OpenAI Jukebox, though. A project abandoned by OpenAI in 2021, but I have recently figured out how to get cleaner sounding results from it and though I couldn't recreate Pavarotti (he's missing from the lists of 5000+ singers/bands) I could recreate Andrea Bocelli singing any combination of words and it wouldn't be hard, just time consuming. I have been doing this with the AI "faux" versions of other recording artists such as Tori Amos, Jim Morrison, etc.

Elizabeth Zharoff doesn't think that singing AI will replace singers and neither do I. I think they can live side-by-side though, at some point in the future.

I DO think, though, that there will eventually be a Coke commercial with singing AI cuz there will probably be a "novelty phase" when singing AI gets really good and everyone will be awed by it for awhile 'til they move on to the next thing to be awed by, occupying about as much time and space in the human psyche as did the hula hoop in the late Fifties: we need to keep in mind that singing AI will just not matter much to most people, while "talking AI" and fake news could affect everyone.

Anyone with an interest in singing AI will enjoy watching this entertaining and informative video. Thanks for uploading it, Marty.

Last edited by Michael Zaneski; 03/21/23 02:20 PM.
Fate doesn't hang on a wrong or right choice
Fortune depends on the tone of your voice

-The Divine Comedy (Neil Hannon)
from the song "Songs of Love"
from the album "Casanova" (1996)