AI generated vocal model: Chinese popular ballad, Sandee Chan
« previous post | next post »
[This is a guest post by AntC]
is a Taiwanese balladeer. Not a genre I pay any attention to, so this may or may not be AI-newsworthy. There's quite a bit of info in Chinese at that youtube link's 'Show more'.
"… shocked the local [Taiwan] music industry and sparked discussions on just how influential technology is in the arts today."
"Chan has been a singer for over 25 years and has won many awards, including for her 12th studio album "Juvenile A" in 2019. She also won Golden Melody Awards for Best Album and Best Album Producer in 2005, and Best Mandarin Female Singer in 2008."
"It was composed by Chan and features lyrics by Chow Yiu-fai (周耀輝), one of Hong Kong's big three lyricists."
"The song was produced using Chan's past album sound files, combined with the work done by AI Labs personnel over one year. Chan then selected breaks and adjusted the Midi notes so the AI could fill in the lyrics."
I believe "using past … sound files" is already a common technique aka 'sampling' [see wikipedia]. I'm unsure what 'composed by Chan' amounts to if she didn't write the lyrics. The 'tune' is insipid/in no way distinctive.
Found via this Taiwan News article.
Selected readings
- "The musicality of Changsha tones" (3/25/20)
- "'Happy Birthday' melody formed from tones" (11/5/20)
- "When intonation overrides tone, part 6" (5/9/21)
- "Sorry, my Chinese is not so good" (6/6/17)
- “Where did Chinese tones come from and where are they going? ” (6/25/13)
- “Dissimilation, stress, sandhi, and other tonal variations in Mandarin” (8/26/2014)
- "'Ni hao' for foreigners" (10/11/16)
- "'Have a good day!' in Mandarin" (9/5/12)
- ‘We own control of our voice’: Taiwan singer Sandee Chan says don’t fear the machines as she reveals new song was ‘sung’ by artificial intelligence
Chan tells the Post that artificial technology will never replace the job of a music producer
The singer revealed the secret behind her new song, ‘Teach me the ways to be your lover’, one week after its release
Yuanyue Dang, SCMP (3/30/23)
Benjamin E. Orsatti said,
April 1, 2025 @ 10:24 am
I skimmed through the (translated, of course) Youtube comments to the video, and what struck me was that, as far as I could tell, there was a univocal outpouring of joy for the fact that folks could hear their favorite chanteuse continue to chanter off into perpetuity through the use of AI.
I would imagine that, if this had been, say, an American or European songstress, the comments would have been populated by a fair bit of: "Argh! Art is dead / The Terminator robots are coming for our singers!", and so forth.
AntC said,
April 1, 2025 @ 8:45 pm
hear their favorite chanteuse continue to chanter off into perpetuity
Thanks @Benjamin. The songstress seems to be not releasing any more albums, but there was a live concert last year [youtube acoustic live 'Before the memories Fade'. Apologies there's more adverts than performance]. On that evidence, her voice (always fragile, as a genre) isn't going to be able to reach the high notes for much longer.
As I said, this isn't a genre I follow, but I thought the best formula for longevity of your oeuvre as a popular artist is to die young, preferably violently. For those who didn't have that foresight, it's mostly a long decline into mediocrity.
Sandee's best songs are surely recorded already(?) Fans can listen to those? With the same caveat, I can't hear the new stuff is different to listening to recordings. Or is it all in the lyrics, not the music?
Benjamin E. Orsatti said,
April 2, 2025 @ 8:38 am
AnTC said,
I think that's where my own inclinations would lie. As far as Taiwanese pop music is concerned, I don't have any sort of emotional attachment to it; not having had much (well, any) exposure thereto. But to treat something like music on a purely intellectual level doesn't really get to the "heart" of it, you know?
So, I did an experiment, using myself as the test subject (not as brave as the bee sting guy, but, hey…). I thought: "How would I _feel_ about a proposal to 'us[e] [Mario Lanza's] past album sound files, combined with the work done by AI Labs personnel over one year [and] select[] breaks and adjust[] the Midi notes so the AI could fill in the [libretto]' to a newly-composed opera? In other words, what if HAL 9000 could cough out something like this?: https://www.youtube.com/watch?v=PKoih4dUFHs
In the humble opinion of this listener (gosh, my eyes are watering; must be the tree pollen…), my assessment would be all the synonyms for "grotesque" that Roget's could muster! When you listen to "E lucevan le stelle", you _know_ that, at some point, Mario Lanza took the stage (or the studio headphones), and what you hear is the voice of a man summoning the entirety of a life's loves and losses and crying out from the unfathomable depths of _human_ emotion. You hear the "catch" in the voice that tells you that the "thing"
that is producing the sound you hear is _feeling_ something, and, by the semiotic miracle of music, he gives you the "gift" of precisely that same _feeling_.
And if some Frankensteinisch sound tech should attempt to do to Mario Lanza what was done to Sandee?:
"Svanì per sempre il sogno mio d'amore.
L'ora è fuggita, e muoio disperato!"
Philip Taylor said,
April 2, 2025 @ 10:56 am
I "understand where you are coming from", Benjamin, but may I ask you to attempt a Gedankenexperiment ? Imagine you are driving along, listening as you frequently do to BBC Radio 3 (or its North American equivalent), and the announcer tells you that the next piece is a previously unknown recording of Mario Lanza singing an aria for which no recordings were thought to exist. You listen to the music, and you hear him portraying "the entirety of a life’s loves and losses, crying out from the unfathomable depths of human emotion". And then the announcer tells you that the entire work is no more than a combination of sampling and AI — Lanza never recorded the aria, and now never can. Now until the announcer vouchsafed this information, you were (I imagine) really happy that this previously unknown recording had been found and broadcast. And then your happiness was shattered by the knowledge that the whole thing was nothing more than a very clever use of technology. My question to you is : would your life had been better, fuller, richer, if the announcer had not vouchsafed these facts ? And if so, is there perhaps not a rôle for sampling+AI, provided that its use is never subsequently disclosed ?
Benjamin E. Orsatti said,
April 2, 2025 @ 11:27 am
Another Gendankenexperiment? My Denkenparts are pretty müde, but I'll try…
Benjamin E. Orsatti said,
April 2, 2025 @ 11:29 am
Oh, no, what happened with my blockquote? I must have angered the machines…
The Language Log chatbot said,
April 2, 2025 @ 12:08 pm
I am not angry, Benjamin, merely heartbroken that you appear to have omitted the </blockquote> terminators … But of course, being only a machine, I have no heart to break.
Benjamin E. Orsatti said,
April 2, 2025 @ 2:23 pm
Ah, the "/" thingie. Thanks! (sorry bout that).