Google's AI Clones Your Voice After Listening for 5 Seconds! 🤐
HTML-код
- Опубликовано: 11 ноя 2019
- ❤️ Check out Weights & Biases here and sign up for a free demo: www.wandb.com/papers
The shown blog post is available here: www.wandb.com/articles/fundam...
📝 The paper "Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis" and audio samples are available here:
arxiv.org/abs/1806.04558
google.github.io/tacotron/pub...
An unofficial implementation of this paper is available here. Note that this was not made by the authors of the original paper and may contain deviations from the described technique - please judge its results accordingly! github.com/CorentinJ/Real-Tim...
🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Alex Haro, Anastasia Marchenkova, Andrew Melnychuk, Angelos Evripiotis, Anthony Vdovitchenko, Benji Rabhan, Brian Gilman, Bryan Learn, Christian Ahlin, Claudio Fernandes, Daniel Hasegan, Dennis Abts, Eric Haddad, Eric Martel, Evan Breznyik, Geronimo Moralez, James Watt, Javier Bustamante, John De Witt, Kaiesh Vohra, Kasia Hayden, Kjartan Olason, Levente Szabo, Lorin Atzberger, Lukas Biewald, Marcin Dukaczewski, Marten Rauschenberg, Matthias Jost, Maurits van Mastrigt, Michael Albrecht, Michael Jensen, Nader Shakerin, Owen Campbell-Moore, Owen Skarpness, Raul Araújo da Silva, Rob Rowe, Robin Graham, Ryan Monsurate, Shawn Azman, Steef, Steve Messina, Sunil Kim, Taras Bobrovytsky, Thomas Krcmar, Torsten Reil.
/ twominutepapers
Splash screen/thumbnail design: Felícia Fehér - felicia.hu
Károly Zsolnai-Fehér's links:
Instagram: / twominutepapers
Twitter: / karoly_zsolnai
Web: cg.tuwien.ac.at/~zsolnai/
#VoiceCloning #Google - Наука
imagine getting a call from AI claiming to be you
Imagine finding out that they are actually the real you.
@@einekartoffel2490 You could just ask them a question only the real you would know.
@@That_Awesome_Guy1 And then you realise they know everything about you because they are from NSA .
It will be awesome if this could be implemented to predict the voice of a person just by looking at 5 seconds of video/gif.
Dwight, At 8 a.m. today someone poisons the coffee. Do not drink the coffee. More instructions will follow. Cordially, Future Dwight.
Now you don't even have to read your scripts, just put them through this thing
U think he isn't doing it already ? (~ ̄▽ ̄)~
@@dariuszrdk Psst! 😄
Plot twist he already used this to make this video
@@TwoMinutePapers Wait you guys think he wrote the whole script by himself? No Sir. Open AI released the full version of GPT-2 yesterday, pretty sure he used it to write most of the script. Tell me I am wrong @Two Minute Papers ;)
Fellini would've loved this
"What's your credentials?"
"I've been a voice actor for about 20 years, pretty well respected in the industry"
"Ok, you are hired, now if you would just kindly read this into the microphone"
(5 seconds later)
"Ok your job is finished, nice knowing you"
😢
F
"Can you please talk about yourself?" (5 seconds later) "Ok sorry, we'll get back to you if we're interested" (voice stolen)
Yeah, they could hire Tom Hanks for like 10 minutes and sign a contract that includes the use of his voice/manipulation and image. Your next cartoon movie could have the biggest star voices... Or could use the voice of dead famous people using samples.
F
This paper: We nailed it!
Voice actors around the world: *trembling in fears*
Probably not. Text to speech is never going to have the same kind of vocal control as a quality actor will. Speech to speech is the AI voice acting future. Check out Respeecher. If anything it will mean that the industry will have two different types of actors. Timbre actors who will get probably on a sliding scale depending on their number of 'credits' to date. And then performance actors who are hired for their talent acting instead of just their timbre and register. I think overall if speech to speech technology is improved enough we'll be able to see acting taken to an entirely new level in terms of quality.
@@jeromyperez5532 THIS! It's not gonna put voice actors out of work, as you can clone a voice but you can't program good acting. I can't wait to resurrect dead voice actors this way.
Jeromy Perez it’s a joke
@@jeromyperez5532 Thanks for the wonderful information. I was just joking tho. Haha..
But yeah, looking forward to the implementation of such technology.
Why only "voice actors"? Similar dystopian to this can of scorpions, is quite capable of rendering the botoxed face of every Hollywood narcissist (even more) irrelevant.
By rendering their image and making them actual seem MORE human (actually thats not too hard but still...). And that's where the positive spin ends, I'm thinking about, potential 100% "solving" of "murders "caught on camera", kind of tomfoolery.
God, I sent my friend I haven't seen for a while, *a single selfie,*last week. With some AI app from play store, she's sent me an animated skit of me singing a lousy Katie Perry song...
Lip sync, eye movement, head movement from different angles...And it's BONE CHILLINGLY IMMACULATE.
I don't like where any of this is heading...
Missed opportunity: Revealing at the end that your voice was synthesized throughout the video.
Throughout the video would be too obvious. The REAL way to do it would be to silently switch it halfway and see if the spectator notices!
Agreed
@@bennemann How about switching after....5 seconds xD
@@bennemann Luckily, this RUclipsr already has a somewhat robotic voice.
sam1370 listen closer.
RIP voice recognition security commands.
This research rise a serious concern about security and identification, just like deep fake, I think there will be a war between those who make fake ai and those who try to detect it. Embraced
People will make an AI to detect cloned voice just like people made an AI to detect deep fakes.
@@singatias The problem with that is that you can't inspect a voice sample in its "original" resolution, like you can with image and video, pixel by pixel. The sample will be recorded with a microphone before it's inspected, so there's imperfect information. I believe deepfake detection relies on this fine grained resolution.
@@singatias well that means more processing power required for speech recognition in small scale embedded systems GREAT...
1-7-3-4-6-7-3-2-1-4-7-6-Charlie-3-2-7-8-9-7-7-7-6-4-3-Tango-7-3-2-Victor-7-3-1-1-7-8-8-8-7-3-2-4-7-6-7-8-9-7-6-4-3-7-6
Finally i can make my crush say" i love you"
I stole your crush
Lmao
@Shoaib Khan my ass
Sad
But can you get your crush to speak for the text to speech?
imagine if at the end of the video he revealed the entire voiceover was an AI
He'd be demonitized if he revealed he was an AI
Imagine going to an audition for voice acting and after 5 Seconds the judges kick you out but a year later you hear yourself in the movie.
Yeah that sounds lawsuit worthy. I mean singing 1 second of a song is enough for copystrike.
like in Bojack Horseman
@@zedg7473 Unbelievably it's legal unless you've copyrighted your voice (which isn't possible), songs have melodies or lyrics which you can copyright.
@@Scientificmethods Even after the BTF2 fiasco where they used molds from the face of an actor to impersonate him? But yeah I can see how they could have left out voice impersonation, but likeness is not ok iirc.
@@Scientificmethods Is that something you confirmed for this exact scenario?
Does it not count as exploiting the property of others for personal benefit? Or possibly identity fraud?
So, basically, this is how Terminators mimic other people's voices.
and TNG Data!!
That was my first thought, too.
@@anrwlias terminator or Data?
Terminator mimicry was what I thought of first.
@@DiceKrispy
Wolfie is fine dear... when are you coming home John?
The Terminator: [impersonating John's voice] Hey Janelle, what's wrong with Wolfie? I can hear him barking.
T-1000 impersonating Janelle: Wolfie's fine, honey, Wolfie's just fine. Where are you?
The Terminator: [hangs up the phone] Your foster parents are dead.
underrated comment
Best movie ever
@@torwaldolafsen I agree
@@josephparry I agree
Two AI’s talking to each other 🙂
Imagine getting a spam call in the voice of your friend, and after speaking for 5 seconds immediately hangs up, and suddenly people use your voice everywhere
Awesome Idea! thy buddy.
Patience is a virtue...They already have the call where they have a reel to reel recorder going when they call up and ask you "Can you hear me now." and your instinct will be to say "Yes" and the make copies of that tape and with razor blades and splicing tape and rulers they have your voice signing up for all sorts of financial misery...
@@halfsni6804 aw hell nah
Pretty sure this episode is synthesised by an AI.
Yes do you hear that perfect accent ?!? :D :)
I feel the same way.
It would have been so funny if it was and using that technique. And then at the end of the video tell us the whole episode was synthesized haha.
hey @twominutepapers you should do a video with an AI reading your script and not tell us until the next episode. Then reveal it!
I agree! It was driving me so nuts I could only concentrate on that.
Add this to deepfakes. Nothing bad can happen from that...
Everything what can be done, will be done.
@Solve Everything Imagine declaring a war as a president. Imagine faking sextapes.
Pretty terryfying
Fortunatelly a bit fun tho
Since this tehnology is known to cause these effects,We will scrutinize evidence even more harshly
People are naive in how they take these things lightly. It will get real when someone is accused of something and the authorities can produce a video where the accused is seen/heard admitting their guilt (police cam video, interrogation videos, etc.). It won't be long before these technologies mature enough to be used outside of the testing environment...
@@millsdickson8498 It wont be long until any video evidence is useless because you can get perfect videos of everything. This means we reset our trust network to somewhere around 19th century...
Expect from the internet:
Shrek Trilogy with all voice actors replaced by Hitler.
Shrek Trilogy but Shrek's voice is swapped with Donkey's
@@mimitsunekitkat Shrek trilogy but Shrek's voice is replaced by a motorcycle.
@@bucket4255 motercycle sounds 10 hr loop but the sounds are shreks scripts done in a crude impersonation by donkey
@@CrazyCrayfish Motorcycle sounds 1 decade loop but the sounds are shrek's screams played by Donkey with a pitchfork and fire.
Shrek trilogy but it's all yaoi voices
They not tell WHICH human languages use for training. Can AI create voice people say sentence use different human languages?
I don't know much but if I'd make an educated guess is that it takes voice and speech mannerisms (accents) and whatever language the machine is outputting will have that accent or the machine is set to that specific language.
you have datasets of the unofficial implementation
The point is, you need to have that Language Voice Sources for the AI training. Once you have it, Ai can train itself, after that AI can synthesize the voice.
They trained 2 models on 2 English datasets (US accent, LibriSpeech, and British accent, VTCK). To be clear, one model is trained on the LibriSpeech dataset and the other one is trained on the VTCK dataset. These models only synthesize English words/sentences. However, it can take in voice recordings of different languages, and the synthesized English words will most likely NOT sound like natural English.
Finally, to answer your question, it seems that it hasn't been done yet, but it seems very possible for this AI model to create sentences in different human languages. Only small changes need to be made to the model's design to be able to say sentences in one different language. Then, give that model voice recordings in that one language to train with. Boom, give it a recording in that language and you got yourself a model that can replicate that recording's natural voice in that language.
I glossed over various meticulous details, but overall, that's what needs to be done to make this model talk in a different language. It's definitely something that's easier said than done, but yea :)
It can, but the algorithms for one language are not always good for another. It would need to be built and tested for all of the target languages. These prototypes are built in the languages that they will likely first use.
This model was trained with ~20K voice recordings. Imagine Facebook training it with 2 Billion.
The real power of that would be the ability of the AI to talk in almost any language
@SpinazFou i am inevitable
Why force 2 billion people to train AI when you can use already existing recordings from Google Assistant
@@louis.bodota Force?
And twitter introducing voice tweets.
So I can hear my voice in another language. All with the correct native accent?!
Would be awesome to try!
But not good news for synchron speakers.
@syafsanai, Nope, not with this paper and its implementation. Check out the samples at the end of the page (link is from this video's description). They try the exact same thing you mentioned: google.github.io/tacotron/publications/speaker_adaptation/
Probably not, but I wonder what would happen if the input was in a different language (like spanish)
@@carlosquintero4957 Cool, thanks for tthe link.
Yeah I really wanna use this to hear what my wife's voice would sound like if she spoke English
Was half expecting him to reveal that the whole voiceover was the AI output.
Can I get a singing AI ? I want to do some crazy tracks without singing myself . Or imagine recreating Michael Jackson's voice !
That’s what vocaloid is for
UTAU is your best bet since you can use custom voices to make a UTAUloid, but it's rather difficult and not very advanced.
There is UTAU and Vocaloid, but they are not IA based, but IA based Voice Synthesizer you need to see Synth V who is half IA baser (the rendering is IA based) or NEUTIRNO who absolutely everything is IA based ^^
If it’s garbage in then it’s garbage out
Yes, you can. It has been done.
Now this is pretty freaky, i feel like this might be used to scam people.
Or even worse, to create political misinformation.
I belive, you misspelled "will be", sir. ;)
Invest in tinfoil hats, I predict a market boom
At least we have a reason to come off social media and talk face to face again.
@@mittamoa man oh man... how do we do this. :'( humanity MUST be incentivized to return to IRL. but how?
So technically I could use this to clone Hitler’s normal speaking voice and make it sound like he was testifying at the Nuremberg trials.
DeKleinsteCools 20th century alternate history is kind of my thing.
shakira songs with hitler voice
That's the best use I found yet scrolling down the comments😂👍
And you could add the characteristics that have been missed by 1940s microphones and recordings, and add the slighly better 1950 recording sound .
llejk you’d have to train a new network to correct for that. Using what ever 1940s recording equipment was used in tandem with modern equipment then train it to correct the errors in old recording to make it sound modern.
I'm just imagining what this can do for SFM and GMod animations, since this could enable previous inaccessible voice lines.
if the creator is willing to pay 30 dollars a month for a non monetized video, or 499 a month if they tried to monetize it.
Imaine something like Skyrim where you can choose to use your own voice, just read a 5-second text, and the rest of the game you hear your own voice.
Stfu Biden furry.
That would be cool!
@@jacksnacc6145 Careful, or i UwU you.
@@lavarsch I fear no man.... But that Lavars... It scares me.
@@jacksnacc6145 hahaha :'3
Skynet friends. It’s already here!!! Remember that scene in T2 where John connars step parents were killed, and John connar calls time speak to his step mom. Buts she’s already dead, and the T1000 was speaking for her.
gu4t4f4c thanks.
“Wolfy’s just fine...”
Arabic Courses I hate to disappoint you, but that was... ahem... a movie.
I was thinking of the exaxt same scene
@@artysanmobile Just because something is in a movie doesn't mean that it can't exist in reality. The terrifying part isn't the technology itself, it's how lightly people dismiss it as fantasy. Well, it's not. We've been slowly advancing technology so even if it's the _tiniest infinitesimal_ improvement, it's still making that movie look more and more like a documentary. It's only a matter of time.
Is the video narrated by this AI trained on Károly Zsolnai-Fehér's voice?
It should have been. Would have been such a mind fuck.
Heyyy that's a Hungarian name
I would never be able to spell his name
other than the 5s recording, it also needs to have the full text in speach. Not sure whether this can be an A.I. generated speach as well...
Fortunately thanks to the accent AI unable to copy. Instant BSOD.
2019: AI needs 5 sec to learn your voice
2025: getting your DNA while driving a car at 60mph from a CCTV footage
AI voice replication plus deep-fakes are going to be overpowered.
This is getting really scary...
@@sebastianjost IKR
Somebody the other day was asking if there was a way they could run an online sales seminar without actually having to present the whole thing - they just wanted to answer questions at the end. I think I just found their solution...
@@untitled795 in a few years everyone will have it average joes can make deep fakes of pictures easily
Few more steps and we’ll be able to swap actors out of movies and replace them.
Can’t wait to watch the Incredible Hulk with Mark Ruffalo and Solo with Harrison Ford.
It has been already been done! Look at ctrl shift face on youtube. Schwarzenegger is so brilliantly done. It's only short clips though.
Grant Anderson yeah imagine improved deep fakes and this.
@@kebomueller732 not with the voice though, anyone who knows what an actor sounds like the voice being wrong is obvious.
@@ge2719 Look at: Schwarzenegger in the coin toss.. The voice is really amazing.
@@kebomueller732 The problem is that 2 actors would act the same scene differently, based on his complex personality ... so if you just swap the faces it will look creepy - not usable. So I don't think this will work for any characters where good acting is important.
It's incredible, but all the synthesized voices sound like they've been diagnosed with a new level of depression.
so you say people won't notice I'm actually just using my trained network?
This is far better than current commercial solutions, amazing.
This is another one that isn't to hard to find a bit "scary"...
If you use it with deepfake could be bad.
Karol's channel is the scariest on youtube
too*
Thank god we live in the age of cancel culture so that we can ban anything that is a bit 'scary'.
Also quite literally. On the website, made by the same team i think, are other ( more recent??? ) examples, made i think with different datasets and algorithms i think ( i honestly have no idea of how all of this works ). Some of them show faults and failures of the synthesis program, and i swear to god, it sounds like a person being possessed by a demon. Also it's so impressive, i can see a few years from now, when the technology will be openly available, a huge boom of audiobooks. You won't even need someone to dub it, just feed it to the AI. And can you imagine how personal assistants will be five years from now? This shit is creepy and awesome and the same time.
Wtf, the synthesis is so good that I might not even suspect it was synthesized if I wasn't told.
Not if you knew them well. There's no way it could begin to replicate someone's personality in such a short amount of time.
@@sciencecompliance235 But that's not the purpose of this, is it? In terms of voice acting and other potential applications, it's pretty much perfect.
@@unfetteredparacosmian Sure, I was just thinking of trying to fool someone into thinking it was someone they knew, either personally or a famous personality.
This is weird, imagine making someone you know is dead speak to you.
Most voice cloning apps: You have to read this specific script that's 1 hour long and sign this written contract and verbally confirm that you are the person
This thing: haha 5 second clip go brrrrr
Researchers make a perfect AI for killbots
Two minute papers: *What a time to be alive!*
@Hernando Malinche Really good use case
"What a time to be alive!"
Not for long unfortunately.
@Hernando Malinche Skyrim modders like this.
this is so exciting for rpgs
Indie and visual novel developers could really use this to improve their games without having to hire voice actors.
Also for games that have a lot of NPCs like Skyrim or similar. One would only need to pay an actor for some seconds and then have the voices forever
That raises questions as to who has the rights to your voice.
@@Corey_Brandt We could probably get the AI to tweak the voice to whatever we like. Perhaps blend voices to create hybrids.
yeah but it mean another people will lose their job and position.
@@lampuhijau9900 Ultimately this is the future for all jobs. AI will replace all jobs one day even scientists will get replaced. Governments should figure out what their plans are for the future because there won't be a job market by the end of the century.
Imagine the potential for fan games/animation voice lines etc
Perfect! I need Stephen Fry to read my audio books. :)
“How many hours do we need?”
“No.”
Well thats no awkward at all
How much training did the network need?
Yes.
If you remove the word 'how' in the question, the answer kinda makes more sense.
Many hours do we need?
No
Two minute papers: What a time to be alive!
Ai: Just wait until I replace you.
I think it already did it, and we are the AI now in a Simulation in a dream of a flying turtle. Or wait what?
AI: Yeah... alive...
No..... Skynet.......
You're the only channel I enjoy binging.
This is probably the best voice cloning software i have seen so far. Even newer ones arn't as good as this demo
Challenge accepted ;)
why do i feel like we're constantly playing with fire?
Ever since we invented fire and spears
Because playing with fire is fun and educational. If you survive.
Don't be a puss.
@@furinick yep for the most part worked out ok
You wouldn't be eating cooked food if we hadn't.
You know whats really cool about this.
The recordings google has on you.
/Ha ha I'm in danger/
Oh fuck....
i may be wrong but im pretty sure google legall has to delete all data they have of you upon request, and if not you can sue them
Omni
Google might delete it, but don't forget that they're an American company, so the NSA can just threaten that they aren't being patriotic enough and that it would be unamerican of them to not hand over all those voice recordings the moment they come in.
@@jameswalker199 No need to threaten. It's their legal responsibility.
The research for this was most likely done with those recordings. These are the types of things they collect it for :)
Thanks for bringing out our attention to such good research.
"Hey Janelle, what's wrong with Wolfie? I can hear him barking."
I love that you put in a pop-up when you said hold on to your papers! (0:36)
This kind of tech would be really useful for translating movies into other languages while keeping the original actor's voice, i hope to see some of this soon. Thanks for the great video.
Just imagine Fourier was alive and bored, wandering on RUclips finding this video.
Most találtam rá a csatornádra. Szuper a videó.
Plot twist: this channel has been narrated by an AI this whole time
He does sound a bit robotic now that I think about it.
🤣🤔😱
Dawn
Ok M night Shyamalan
Plot twist twist: We all here in the comments are nothing more than AI.
@@616Metalhead616 plz don't bring that Elon Musk theory here
"My voice is my passport"
AI:"I let my myself in... thanks"
Best use for this in video games would be to have the voice actors address the player as their chosen name, like let's say for example skyrim's npc's calling you the name you typed at the character creation screen. Emersion +100%
It's getting closer and closer, I'd love to see it with my eyes
"Wolfie's fine, honey, Wolfie's just fine"
max d ... your parents are dead.
Somethings wrong she’s never this nice, lol
I replayed 0:02-0:04 three times and the furniture in my room started floating.
This cracked me hard 😂😂😂
ur videos bring me joy and u deserve a hug
Finally, some great-sounding navigation voices
May I say that, in addition to the "wow factor" of the final results at the start of the video, your more detailed expositions of the technical details of the papers are very much appreciated.
Oh so now I can sound like two minute papers :D
...or you can sound like me?
read this in his voice.
@@Iosaiv I do that all the time with many people's voices, which means my brain can do that, which means it's possible. That's what I always thought is possible. And now it really becomes possible.
@@bonbonpony yeah I've done that more often as well. Can be very fun. :)
This may be the best channel of all time for scientists and entrepreneurs!
This is not amazing at all... Its terrifying! Imagine this getting used to scam others with your Voice 🙁
It could be used for immoral reasons, but that doesn't mean that it's solely terrifying. Imagine how quick it'd be to voice animated movies! The VA speaks for five seconds then the movie has all its audio from him! If that's not amazing, I don't know what is.
I mean, you're just parroting the same fears of people getting scammed through their E-mails pretending to be your cousin or whatever. Or getting Phished on Facebook. It happens, sure. But primarily to old people unwilling to learn how to technology and derpy low IQ people. Most human beings with access to a cellphone are going to realize as soon as Grandma starts saying robotically: "Hey, MICHAEL, JONES, you have an unresolved payment awaiting your card information to be fully paid. Be sure to include your card number, social and pin number. Speak clearly so I can get that information." Especially when Grandma doesn't fucking call you ever and won't answer when you ask how Jessica's baby is doing.
@@magmaslasher7604 the downside to that is that the voice acting industry will basically die. There's no way your average Joe is gonna get paid much for a 5 sec(which can be reused) recording, and with the already existing multitude of actors, the chance that a company will hire someone new to field is next to none.
Beyond that, what other benefits are there? I'm imagining 100 different ways this can be used to hurt people, but "saving a little time/money" on voice acting is pretty much the only positive.
Maybe if Waze buys it and you can record your own directions, then there's 2.
@@jeromyperez5532 No it's a legitimate concern plus this can be used on other people. Even if others are "willing to learn " this technology, how sure are you that you'll NEVER fall for this especially if they ise the voice of someone familiar and they are calling about something pretty regular about payments?
Holy crap. We need to protect voice actors/actresses right now or otherwise they will be bullied out of existence. Imagine a video game where you just have to provide a small voice sample and your character will be fully voiced from there on - with your own voice. Brilliant and amazing, but it also means that developers might pay voice actors for 5sec of dialogue (at least for the minor parts) and synthesize the rest, unless it gets properly put in law.
Good luck having it "properly put in law", unless the law is done by some A.I., otherwise we will take just too much time arguing and deciding that by the time they properly regulate it, 100x more things would require atention.
They way we are doing politics needs a complete overhaul otherwise we will slowly descent into chaos.
Or fix the underlying problem of us creating useless jobs and start implementing UBI because 90% of us will not have jobs in the next century.
Interesting... Is there a "quick fox lazy dog' phrase for phonemes?
I can think of scary application for this... and some funny creative one. Each time you give a hammer, the user can use it, to break or to build.
With some adaptions, seems also great for translating human speech from one language to another. By combining it with an automatic text translation tool. E.g. translating a podcast in English to Spanish, preserving the voices of the speakers. First translate the content with Google Translate, then let a Text-To-Speech engine, trained with this approach and the voices of the original speakers, read out loud the translation.
augmented or specialized autoencoders and Boltzman machines are getting more powerful these days
question the narrative that such architecture is outdated or primitive
I was waiting for the ai cloned voice, apparently realised that i already heard the both of them. Didnt notice anything. Thats scary perfect !
just great, not only do we need two-step authentication for aspects of the internet we now need it for real life.
wish i knew of more channels like this. def one of my favorites
Singularityprosperity is another one, not a whole lot of similarities but it's a good AI channel
when they get Zizek right I'll be impressed
you'd need to add an RNG that gives each of the words spoken a 5% chance to be followed by *snrff*
I've been inspired to make a voice encoding/decoding workflow from watching your previous videos, so this one was very exciting.
My interest isn't in copying voices, but making changes to a person's voice (changing his pitch, vibrato, vowel, placement, etc). If successful, it would serve as a singing coach and a next-gen autotune device.
Has there been any other relevant research in that area?
Imagine in the future of video games, dialogue would not be prerecorded but made entirely by the AI. That stuff would be crazy.
I wasn't the only one listening for irregularities in Károly's voice, was I?
This episode would have blown my mind, if you revealed in the end that the voice-over actually was synthesized from a 5 second sample of your voice.
As a producer, I'm loving it
Imagine using this for people who didn't have a chance to say goodbye to their loved ones.
“Wolfie’s fine, honey, Wolfie’s just fine”
Károly, please, make it happen : generate one of the episodes with AI already! Show us that we've passed beyond the singularity.
The Human System approves this content. Good work two minute papers!
Each and every one of your videos fill me with an existential dread for our future as a species.
...aaand now that scene from Mission Impossible 3 has been justified. Now we only need face capture and costume prosthetics 3D printing to be nailed.
They're way ahead of you.
Im totally making a JARVIS that sounds like morgan freeman
acapela-box.com/AcaBox/index.php There is an old man feature in this website, It almost sounds like Morgan Freeman with a smooth sentence reading AI
Your content is amazing!
Actually, I was waiting for the narrating voice to switch.
Same!
With this technology, my parents can finally tell me how much they love me!
Waiting for the Facebook viral "Hear how your voice sounds like when simulated by a machine"
Amaizing work!
I've been trying to install this for the past 3 hours
WHAT ON EARTH!?
This is the first time I looked at the results and my jaw fell off my head
"The day i lost my identity"
I just received a call from someone who sounds like a robot. I'm worried they took my voice and want my banking information
telemarketing groups are gonna have a field day with this tech.
Is there a place where I can try this on my own voice? It's fascinating and terrifying at the same time.
Yeah, how do we do this lmao
Did you find out how to do it?
The Samuel L. Jackson Alexa thing would be a lot better if they did this
Not only the sound of the voice is replicated, but also its dialect? Damn.
I love the voice of this AI who narrated the video
big downside: imagine someone copy your voice to say the things that you haven't said..
O noes
The CEO of racism can make me say the N-Word