What if You Could Sing in Your Favorite Musician's Voice? | Holly Herndon | TED
HTML-код
- Опубликовано: 28 авг 2024
- What if you could create new music using your favorite musician's voice? Sharing her melodic gifts with the world, multidisciplinary artist Holly Herndon introduces Holly+, an AI-powered instrument that lets people sing with her own voice. Musician Pher joins her onstage to demonstrate this mind-blowing tech while singing into two microphones -- one that amplifies his natural voice and another that makes him sound just like Holly.
If you love watching TED Talks like this one, become a TED Member to support our mission of spreading ideas: ted.com/membership
Follow TED!
Twitter: / tedtalks
Instagram: / ted
Facebook: / ted
LinkedIn: / ted-conferences
TikTok: / tedtoks
The TED Talks channel features talks, performances and original series from the world's leading thinkers and doers. Subscribe to our channel for videos on Technology, Entertainment and Design - plus science, business, global issues, the arts and more. Visit TED.com to get our entire library of TED Talks, transcripts, translations, personalized talk recommendations and more.
Watch more: go.ted.com/hol...
• What if You Could Sing...
TED's videos may be used for non-commercial purposes under a Creative Commons License, Attribution-Non Commercial-No Derivatives (or the CC BY - NC - ND 4.0 International) and in accordance with our TED Talks Usage Policy (www.ted.com/ab.... For more information on using TED for commercial purposes (e.g. employee learning, in a film or online course), please submit a Media Request at media-requests...
#TED #TEDTalks #Holly+
And the solo duet is born...
Simultanious*
I'm always impressed by people who have the endless curiosity and drive to keep investigating the artistic possibilities of new technological developments. Holly Herndon is the epitome of this approach, and gets my enduring respect.
The dangerous thing is that in Latin America they could use this technology to scam by telephone and thus steal
Holly plus does not work. After watching this TED talk, i went to try it right away and the audio I supplied was a simple Hindi audio which could not transform a single word... The output was crap garbled audio ... Tested with multiple settings provided on the website... Waste
The audio provided and demonstrated in this video is not real for sure after my tests
necesitamos la versión completa de bésame mucho, holly por favor
Karaoke night just got a lot better.
"I'd like to sing a Another One Bite The Dust in Barry White's voice, please."
Karaoke voice clones have been a pretty normal thing in East Asia for a while now, with companies like Yamaha working on this stuff for years. Check out Yamaha's "TransVOX" for just one example.
Scary to imagine how this could be used the wrong way
This is wild! The immortalization of a voice could have a lot of practical use for those with vocal disabilities who lose access to their voice, giving them the ability to sing again (or even the first time!) and make musical careers more accessible to them. Or maybe even beloved, aging musicians who love to perform but begin to lose their vocal strength over time. I wonder if it's possible to put an auditory equivalent of a watermark on the tracks so you could tell if this AI was being used.
It’s unsettling. I’ve worked my entire life to get the voice I have. I just don’t know how to feel about this
That is a small minority of cases. It damns more people that it empowers.
Potential challenges:
-production companies owning someone’s voice forever
-people potentially stealing someone’s voice. (Literally look at 2d art space now, it’s just openly used more for theft and scams. And it’s only a few months old)
-harder for people who put in effort to actually get work, and if they do they could have it wholesale taken from them
It sounds horrible but maybe it’s better off that way when the majority living in a weird dystopia. Musicians and Artists already struggle as is.
The future is coming 🎉
It's unfair to those who worked so hard to get the voices they did, and it's also unfair to those disabled who have learned to live with their disability and adjusted to it.
Voctrolabs announcing this Holly+ and months later Yamaha launches Vocaloid 6 with Vocalochanger that brings the same function, remembering that Vocaloid is a creation of Voctrolabs with Yamaha
Caveat Listener - any speaker who credits ‘Mack The Knife’ only to ‘Brecht’ (1:09)
is not someone to be taken seriously. Kurt Weill wrote the music you hear;
Brecht was the lyric writer/story collaborator with the music composer Weill. (more)
As a professional musician and composer with decades of experience with new technology,
I am always intrigued when ‘new’ things show up in the field of music.
This TED talk presentation is similar to many innovations I’ve seen in the past;
on the surface very impressive, but practical use is much less promising.
Holly is like many other people that hail from the computer/software field -
eager to ‘change the world of music’ with their technological innovation.
However, the ‘smoke and mirrors’ alerts go off when 1)she does not sing anything herself -
2) the amount of processing ( digital delay, reverb, etc.) on the ’spawned’ vocals
3) the glossing over the actual software algorithms created by another party
that do all the hard work of analysis of ‘source’ material.
The comments I’ve seen from those that have tried ‘Holly+’ are that it does not work..
Still, with input from actual musicians, an increase in processing power and AI capabilities,
this early version of vocal replacement technology will be considered
at the stage in development not unlike the Apple Newton -
a good idea, poorly designed and implemented,
that was the precursor to what eventually became a useful tool - the iPhone.
I suggest singers work on their craft like Pher obviously has -
it is his human voice (tone, phrasing, expression) that makes any of this worth listening to.
Her and Grimes, a collab 💛
I love art and everything but this is so marketable. Imagine the karaoke bars with this technology and a whole catalog of all the best artists
It would ruin karaoke tbh. If everyone can sing perfectly then beauty will die
I don't know if others noticed this, but his vibrato made her voice sound out of tune sometimes, even when on Holly+ alone. "a wide variety of phrases in my entire vocal range" is too vague if she meant musical phrases or spoken. If musical it could be clashing vibratos, if not it would be so much more interesting to understand why.
It could even be just throwing the singer off when it changed.
Well, whatever imperfections you hear will be worked out. This is absolutely crazy what they are doing
@@nomandad2000 I just thought it was interesting that's happening, I wasn't criticizing it...
I heard that at times the AI voice was slightly out of tune too…
What about phrasing, mood, breath, and
sweat in the music
Amazing and scary what we can do with technology. You can hear a slight lag where the AI is taking the song from the Live mic and outputting the female voice. If they could get that down to 0 then it'd be pretty near flawless. But it's still scary knowing that deepfaking is at the point where we can make audio and video of someone that is completely manufactured and fake. Think about that for a minute. Then remember that a couple of students at MIT developed Deepfake technology back in the 90's and it's taken until now for the rest of us to have access to it.
I’d say the results would not be as mind-blowing if it didn’t have a buffer. Pitch detection needs an amount of data to do it well. This is the same in autotune, harmonizers etc.
...or you could alternatively just delay the live input of his mic. all modern digital consoles can and do do that, on purpose for certain applications. virtually every signal you hear out of a PA has digital latency from conversion and fx and so on.
i wonder whats in his iems though. does he listen to his own voice, holly± or both 😂
pretty cool...how is there no latency, is it actually processing or just getting a gate from the microphone and following pre written lyrics?...guess I'll have to look into it
There is a bit of latency compared to Pher's signal if you pay attention (100-200 ms?)
Very little latency for what it is doing.
I make my living in a studio using samples as well as recording real singers and instruments and I'm impressed. Great singing there too. Love it!
Excellent, very good sound and very interesting content. Can be used to learn English for me who can't yet hehe I hope the content continues to grow, cheers
With this you can create a fully virtual singer, is fucking insane
_Miku says hi._
Sound could rappers down bad right now
@@SolPhantasmagoria Miku needs human manipulation in order for her to sing. He can't do anything without someone handling his voice.
Love the Herndon
Beatboxers need to work with this tech
I went to a conference on AI in music and got to have a go on a model trained on Giorgos Gargalas' beatboxing style. Lots of fun to hear your own silly mouth noises come back at you as a professional beatboxer. Unfortunately I can't find any vids of this specific AI demo online. It was really cool though
omg. pher and holly+ what a bizarre, delightful, spellbinding pair
Amazing but scary stuff!!!
Imaging training Holy+ with Frank Sinatra's, Michael Jackson's, Elvis Presley's voice????
Yes, it's a scary thing on the horizon. We will see that soon.
Fantastic and at the same time pose ethic issues that we should consider…
This is so much better than any other voice changer out there. How come we don't have access to stuff like this?
I'm catalan and i love this intro but it was difficult to understand the language tbh
That’s why Spain’s in charge
@@hdmat101 ?
Same with the German song, even knowing the lyrics. Still very fascinating
Ive been waiting for this technology
this is so exciting!
I wish I could use Nina Simone's voice. Will it be possible to create a Nina+ or Kurt+ with just the recordings that we have ?
I believe it is possible once the vocal range, voice texture and intonations of an artist have been sampled and correctly tweaked into the A.I. model.
I am catalan and I barely managed to understand one word of the opening song
Probably because catalan is a rarely used language and the training in it was not good, I understand Spanish and the part sung in Spanish was perfect.
a beautiful mind indeed. wonderful creative innovation.
I would like to sing in the voice of Yo-Yo Ma’s cello.
That was amazing
How’s Wolfie? I can hear him barking.
The live singing with her voice would be difficult, when singing, you rely on the familiarity of hearing your own voice, try singing without hearing yourself, or being on stage and you cant hear yourself in the mix, its very hard! In 10 years time, you will be able to ask Spotify to change the singer on what you are listening to, or create a song in the style of someone with a singer with aspects of whoever you want. I'll finally be able to sing Back in Black and stop telling the audience I can only sing the Bon Scott era AC/DC songs!
Until you get adjusted to a new voice😅😅😅😅
singers have been using in ear monitoring for quite a while now, for various reasons. you can still listen to your own voice over the music.
if you only heard the timbre transfered signal, that might get a little wild
@@recordednowhere I have IEM's and I am a singer, trying to sing live with someones else voice, would not work, other than as novelty act. However, for large stadium shows, often they blend the live performance in with recordings and autotune effected out signals, or simply play a midi triggered recording and mine along, this would work, those kind of shows are fake anyway!
@@makinganoise6028what i mean is, there's no need to hear the other voice as a singer. i think this works best if you just have your own voice in the iems and the other voice gets sent out to the listeners, via pa or livestream or whatever.
Incredible.
Just be yourself..
Music in its true self is simply you enjoying vibrations.
No training or conformity needed.
The imperfection of humans is what makes art and music enjoyable... Perfect pitch and no natural variations in timing sounds homogenized...
Legit lmao i would listen to something human any day over this. Also this woman creeps me the heck out.
Ai-dolatry is simply scary. You will lose your own identity to a lifeless technology. Machines should help humans not replace them. Ai is... Altered identity
I would have named the AI "Ursula"
😂😂😂 right
WOW! This is amazing....
This is amazing and soulful
Direct face to face communication will be humankind's only salvation.
who is your coach if this is not too much i'm asking? I've been looking into advisors lately myself, my retirement plans are going down the drain, my 401k has particularly lost everything gained since 2019
@RichardTrade219
Hey - I've been hearing about Richard for a long time now. Started during the pandemic and I've heard of how accurate his signals is.. I really do appreciate hearing your advice and feel that it is genuine
Holders remains indecisive if this is a continuous uptrend and wether btc would hit it's all time high in mid march as predicted by analysts or if this little uptrend is a trick to a new fallout in bitcoin prices. More the less, now is the perfect time to start trading your coins. I’m not bothered because I’ve made over 16 btc
Some of the Holly+ high notes are flat. I assume that Voctro Labs/Voicemod can fix that.
Wow. So very dangerous. I see scary things in our future.
It is also terrifying how only the ultra affluent and wealthy have accent to this technology first thus determining our digital futures for us without our consent. The next “pioneers” of tech oligarchs are the children of the ultra wealthy techno monarchs
Imagine DMCA striking everything... ;)
This is brilliant 👏
We just got a perfect teacher.
Amazing!!
Perfect 🥰 😍
I even clapping in front of my laptop while watching this alone ! =))
where can we get the software?
beautiful music..made me sad tho...i could feel the spirit coming from him when he was singing...in our future, we will be confused to if it was really him...enjoy what we heard here because it was half way...real.
*BEST USAGE*
*_Picture a scenario where a person like Mr. Beast who has to hire voice actors to translate his videos to other languages decides to use this technology instead. That way his voice still remains in the videos, but appears like he suddenly learnt a new language, and also the voice will be insync with the video, unlike the current dubbed versions. IT would be a game changer!_*
Mr who?
@@Martin-bx1et Google
lol this is such a narrow view to unilaterally declare the BEST USAGE
@@seanplew6548 Haha, I know, I just thought that would be relatable. But I'm starting to think followers of TED are not exactly his kind of followers.
I also came up with this idea, but for movie voice dubbing. It will not be lip synched, though. If you say gun, the AI says pistola, there's no way to lip-synch that without further deepfaking.
It will be cheaper than hiring a professional actor in every sense, but it's good for youtubers and indie gamedevs.
It was only a matter of time for this to happen. Interesting to see the music generated from all the dead singers.
Awesome!
I am ready to hear Nick Cave cover Cat Stevens now.
Ted talk discovered utau and vocaloid
Hehe funny, but AFAIK utau and vocaloid are voice synthesizers, not transformers…
mind blowing
Is there a tutorial on how to use this somewhere?
how amazing
Great!
the way she predicted ai covers
I tried Holly+ with Wesley Willis "Rock N Roll McDonalds" (after separating voice and music) and the result was not convincing
Oh that’s wild!
Synthesizer V AI and CeVIO AI already exist as affordable and easy to use singers. This is not remotely new or special technology.
Also Synth V AI can sing in English, Japanese, and Chinese.
@@kitten48484 I don't think you understood the lecture..
@@1998Cebola I admit I commented for a different thread by mistake, I was supposed to be talking to someone and I lost the comment. Sorry.
Well I still play a double bass made of a tree with a bow made out of wood and horse hair. I will not change what I fell for nothing in this world.
"Mack The Knife, by Brecht"? Oy Weill.
Rastafary said it “I and AI” ❤
👏👏👏👏🌈🦄 - I honestly don't have any words because my jaw in still on the ground!
"Mack The Knife, by Brecht"? Oy Weill
Does anyone know what the web address is for Voctro Labs?
amazing！
It's gorgeous and scary
does anybody know how i do this?
Ethan Hunt liked this.
vielleicht endlich das Ende von Synchronsprechern und Untertiteln
This is crazy!
Wow, holy holly+
Insane 👏👏👏👏👏
Which company is producing this technology?
never before heard sounds
I wonder about this every time I list to my favorite songs
With this technology, we can create songs without a singer, nice, finally the song writer and lyric writer can earn their fair share….
song write on gpt3...
Nothing is real anymore.
Interesting!
Oh great.. First auto-tune, now this.. Why do we even need human talent at all (other than programming skills). I may be old school, but I really dislike this.
Only in a world with basic income and regulations against corporations copyrighting / patenting everything can this be valuable.
Whoa that's cool
Sounds like a holly vocaloid
wow
Ya me imagino a Holly cantando una canción de peso pluma jajaja
I want Freddy back! finally i will be able to be in a real queen concert...
Marc Martel can do that without the tech
@@makinganoise6028 I mean yeah, but there is something about this cuz its literally Freddy digital voice trained with the original voice so ...I know there was always ppl who imitate singers, this is different
I want one of those to synth my farts.
🤣 🤣 🤣 🤣
why ????
I would like to try singing with the voice of Joakim Bróden, ngl...
Lemme sing like Maynard James Keenan and I will sell my soul...again
Is that Hime hair cut?
Wow
So you can make me sound like Corpsegrinder?!
Damnnnnnn