for others wondering why lily got the point pre-tie breaker despite the number being 4- it’s most likely the editor(s)’s fault because lily did say 1, both verbally and with her finger 👍 it’s a bit hard to hear considering that they all said it tgt but yeah,,
i can see why people got sykkuno wrong a lot, the ai was rlly similar to his actual voice but sykkuno kinda has a little fade out at the end of his sentences that the ai couldnt replicate and he just sounded like a regular guy with the ai
Most of the voices are distinguishable by how clearly the syntax of the voices are put in. AI usually can get emotion and such into its generated voices. It has trouble getting the correct spacing between words since everyone usually has their own unique word spacing depending on how comfortable they are, fluency in language, and their current emotion.
some of these were crazy close, it'd be kind of interesting to see this again but with the actual person to see if they could tell their own voice apart lol
I got the sykkuno ones right because he doesn't talk fast, he talks slow with pauses, he drags out his words than saying them quickly and I don't even watch his stream I've just seen him in videos of other content creators XD
@@JasperPeters Issue, AI requires a box to run. AI learning is essentially giving a computer a box and asking it to guess what goes into the box, similar to the green door riddle.
Good game, Toast. Always felt like your true calling was a game master. You and lud need to co-create a gameshow. Lud has an eye for presentation you have an eye for game mechanics
This was so easy, the ai voices are at a lower sampling rate so it is really easy to tell the difference if you know what to listen for The easiest way to describe it is as a slight tearing/static like sound
Oh wow, I didn't think of that but now that you say it you're right. You can also just listen for the high cut, the fakes are missing some high frequencies because of the sample rate.
This was very hard and I don't know any of these people too well but to me all the robot Sykkunos sounded way more "regular guy" and not enough "Sykkuno" so I got them right, and they were the only ones I was confident about.
As far as I know as a AI engineer, the voice AI was trained using a lot of speech data. The AI already "know" how speak like human and how to pronounce words. So the 30 mnt audio input is just to "copy" the speaking style
I wonder if it’s easier as a consistent viewer vs a friend because the sample recordings likely came from stream. As a viewer you only hear their on-stream, usually at least a bit performative, exclusively digital-audio voice vs a friend who hears their voice in so many other contexts. I was surprised I got all of Rae’s but it was like I could literally picture if it was the way it would sound on stream
Very interesting. I kinda wanna know how the AI creates the fake sentences/phrases. Some of the fake ones sounded so real to me thats crazy. Great vid 👍
need more of toast. man, you make life so much better atm. always a good video when you are in it. i pray ur doing well. this is ur sign to go harder than ever *when possible of course*. i really think people are starting to see how good of a content creator you are. go check your views, i bet they're climbing!!!! new vid next week???????????
small editing error at 26:47 - lily guessed 1 but it says 4 whenever toast goes on a rant about how obvious an answer is only to get it wrong, i can feel another year added to my life fr
One thing I noticed is that AI can't reproduce echo in the recording, all the real ones have some echo and are little muffled, where as fake ones are all very clear
I binge watched all Sykuno Decive Inc vods this week after Rae posted video of this game so I got all of him right. I dont watch a lot of John and Sydney because Valorant isnt my game so they were the hardest to guesss. It really depends of how much you hear theirs real voice
Once I realized the AI leaves out end of sentence breaths that people have, it became easy to tell even for creators who I never watched. I think once it can learn little mannerisms like that, then it's all over
i got all peter (since we didn't know what the last one is about, i'll say i at least chose an audio that is fake), i got the last two of john and sydney, and got celine i failed to recognize rae, yvonne and sykkuno, and i have probably seen more of them than john, syd and celine lol
the AI versions are really distinct, it's the night and day difference between the scripted expected way a person would say a line and the real way that a person produces speech. the ai samples don't have the same range of volume or pitch
Check out Dead Island 2 on the Epic Games Store Now!!! epic.gm/DisguisedToast
Wow content disguised 😊
Oh snap right Dead Island 2 is out!
🟢
There was a few times people didn't get points when they got it correct so there is no true winner😂😢 therefore by default toast wins🎉
OH FUCK, TERMINAtor IS JUST AROUND THE CORNER!
Imagine if this video was entirely AI generated and it turned out Toast was really a "Disguised Toast" and never real to begin with
Biggest plot twist since star wars
The assassins are heading to your house right now, you've uncovered too much
Still wouldn't be nearly as shocking as watching a video without your comment below it....
Imagine 💀
maybe it wasnt the disguised toast, but the friends we made all along
Funny how they went from “robots can’t replicate his voice…or his charm” to “that bot almost has Sykkuno down, he’s the first to go” 😂😂
20:16
Sean saying "I hope it's old sykkuno" then the throwback pic of sykkuno appearing had me 💀💀💀
They did mah boy dirty
Naa tht was ai generated
Surely
Otv editors please! ❤
;
i need a part 2 where one of the people guessing has their own question. would be insane is you guys couldn't recognise your own real or fake voices.
It'd be tough since you hear your voice differently than everyone else.
this needds to be pin!! to have them guess on their own voice!
honestly i was expecting that as the final round.
Yes, yes, and yes! I totally agree!
yall are forgetting they would have been the one vouce the line so they would know
I wonder how well they would do finding the AI for their own voices
such a good idea!
I think it should be harder to recognize your own voice
Scarra is a sprout, confirmed.
Lily is the best friend.
this whole thing just made me realise how much of rae I watch everyday 💀
Me too, the first one made me cackle, “guys lets react to stuff.” Lmao rae would never say that
@@romdom9406 im surprised toast got it wrong cause on the 4th one you can hear Mika barking
I mean, the way Rae speaks is also very unique. The AI Raes sound human, but just not like Rae. The cadence and intonation was way off.
@@romdom9406 everything theyve said is stuff theyve said for real before apparently, its just ai copying it
fr lmao
I was hoping we’d get to the “spot the real voice” for Celine and it would just be her screaming 4 times 😂
John's cadence is so unique
for others wondering why lily got the point pre-tie breaker despite the number being 4- it’s most likely the editor(s)’s fault because lily did say 1, both verbally and with her finger 👍 it’s a bit hard to hear considering that they all said it tgt but yeah,,
Thank you because that legit bothered me.
Thanks. I was a bit excited there
Rae’s was by far the easiest, but damn as a whole this was so hard! I struggled so much for sykkuno’s despite watching him a lot lol
i can see why people got sykkuno wrong a lot, the ai was rlly similar to his actual voice
but sykkuno kinda has a little fade out at the end of his sentences that the ai couldnt replicate and he just sounded like a regular guy with the ai
So when is Toast gonna use AI to take a break?
Maybe he is using it already...
Already has
he's just as monotone as a robot, that if he was right now, we'd never know
Wait, is this toast real?
@@BrunoRafaBR no he used an ai for all 4 of them for this video
Lily talking with that attitude to Toast is awesome lol 😂
I like how its super obvious Toast reads off a script for the ad/sponsor, meanwhile you have UA-camrs who cleverly transition to it with a pun or joke
Most of the voices are distinguishable by how clearly the syntax of the voices are put in. AI usually can get emotion and such into its generated voices. It has trouble getting the correct spacing between words since everyone usually has their own unique word spacing depending on how comfortable they are, fluency in language, and their current emotion.
Honestly i got sykkuno all right bc he has that tone in his voice which makes him sound so incredibly nice if he doesnt its not him
Sykkuno is not surviving the robot uprising 😂
some of these were crazy close, it'd be kind of interesting to see this again but with the actual person to see if they could tell their own voice apart lol
I got the sykkuno ones right because he doesn't talk fast, he talks slow with pauses, he drags out his words than saying them quickly and I don't even watch his stream I've just seen him in videos of other content creators XD
This is a genius game show Sir Toast. NEED MORE
Voice actress wins. Checks out. How did Toast get fooled SO HARD
No one can replace Sykunno , his thinking is too outaide the box
💦🔫
Just don't give AI a box, and it'll be able to think just as out the box as syk.
@@JasperPeters Issue, AI requires a box to run. AI learning is essentially giving a computer a box and asking it to guess what goes into the box, similar to the green door riddle.
@@MarkoIronFist Your explanation with the profile pick has me questioning how good AIs are at comments.
@@hulkaiden6178 You cannot distinguish us from the average UA-cam commenter. ◉‿◉
Good game, Toast. Always felt like your true calling was a game master. You and lud need to co-create a gameshow. Lud has an eye for presentation you have an eye for game mechanics
I have a sneaking suspicion once Toast knew he wasn’t gonna win he decided to keep his score at 1 for the memes
This was so easy, the ai voices are at a lower sampling rate so it is really easy to tell the difference if you know what to listen for
The easiest way to describe it is as a slight tearing/static like sound
Oh wow, I didn't think of that but now that you say it you're right. You can also just listen for the high cut, the fakes are missing some high frequencies because of the sample rate.
In the Rae clips the AI wasn't able to replicate the room echo so it made the fake ones very obvious
😂 You did Sykunno dirty by adding that teen photo of him,
This was very hard and I don't know any of these people too well but to me all the robot Sykkunos sounded way more "regular guy" and not enough "Sykkuno" so I got them right, and they were the only ones I was confident about.
Does anyone know what method they used to make these AI voices? Toast mentioned needing only 30 minutes of audio, that seems pretty impressive.
As far as I know as a AI engineer, the voice AI was trained using a lot of speech data. The AI already "know" how speak like human and how to pronounce words. So the 30 mnt audio input is just to "copy" the speaking style
Elevenlabs would be the easiest one.
LMAO NOT THE SPIGEN CASE FOR THE SAMSUNG S5 PICTURE Sykkuno 🤣🤣
i didnt watch sykkuno back then but i had that exact same case LOL
I thought the last one for sykkuno would of been "oh jesus" 😂
I wonder if it’s easier as a consistent viewer vs a friend because the sample recordings likely came from stream. As a viewer you only hear their on-stream, usually at least a bit performative, exclusively digital-audio voice vs a friend who hears their voice in so many other contexts. I was surprised I got all of Rae’s but it was like I could literally picture if it was the way it would sound on stream
All I learned is that I watch Rae way too much 😂
SAME LMAO I GOT ALL OF RAES
Rae has the least monotonous voice
Me too.
Very interesting. I kinda wanna know how the AI creates the fake sentences/phrases. Some of the fake ones sounded so real to me thats crazy. Great vid 👍
The bot is good at imitating Sykkuno bc its also a Sprout
Should’ve done “oh Jesus,” for Sykkuno. 😂
As a Sykkuno fan, I also got the guesses where only Scarra was correct. Syk's voice is kinda airy and sort of comes from his chest.
21:18 literally everyone is pogging
This was a fun idea and well executed. I thoroughly enjoyed
I'm proud I guessed sykkuno's voice correctly
Ngl the only one I knew for sure in my bones was for finding the real one out of fakes for Sykkuno
missed opportunity that Sykunno's third one wasn't "oh Jesus"
I thought was the greatest Sykkuno fan… Scarra proved me wrong.
Id say Sponsored Toast is an AI, but I know Toast would never risk his sponsor.
it's hilarious that scarra knows sykkuno's voice so intimately but asked him to move out the otv house because he wasnt in otv yet didnt sign him
For the people wondering the name of the song in the sponcor clip is "Call of the North" by Niklas Johansson
Omg, an Amongus shapeshifter with AI voice options would be amazing.
Scarra’s sykkuno rounds were gamer god mode, i swear they sounded way to similar, and lily and toast definitely know sykkuno more
I love the Metroid Prime 1 puzzle music playing in the background
Scarra the sprout! He shouldn't let anyone tell him he isn't a real one.
This is such a big brain play idea
need more of toast. man, you make life so much better atm. always a good video when you are in it. i pray ur doing well. this is ur sign to go harder than ever *when possible of course*. i really think people are starting to see how good of a content creator you are. go check your views, i bet they're climbing!!!! new vid next week???????????
I guessed it all in Sykkuno's turn yaaah
Tbf sykkuno speaks like an AI all the time
To be fair. The “muffled” effect makes it a bit harder. Without effects it’s easier
small editing error at 26:47 - lily guessed 1 but it says 4
whenever toast goes on a rant about how obvious an answer is only to get it wrong, i can feel another year added to my life fr
Scarra clapping at 23:25 is so cute 😭
Toast making Turing Test into a gameshow is what we expect for Toast's content
One thing I noticed is that AI can't reproduce echo in the recording, all the real ones have some echo and are little muffled, where as fake ones are all very clear
What I’ve learned today: Scarra is the biggest Sykkuno fan. That’s why he’s the cleaner for the Don
This would be a crazy concept with the individuals like AI toast AI lily etc
The way I got all the Sykkuno ones right bc I had a whole phase with him 😭
I binge watched all Sykuno Decive Inc vods this week after Rae posted video of this game so I got all of him right. I dont watch a lot of John and Sydney because Valorant isnt my game so they were the hardest to guesss. It really depends of how much you hear theirs real voice
This was entertaining to watch. I'd watch a few more of these. Thanks for content to watch and stuff.
Scarra, the President of Sprouts
Scarra swept the sykkuno round
if the professional voice actor didn't win this game show i would have been very disappointed
Not toast picking the ones that sound the most human 😂
18:43 they replayed 1 not 3
For Sydney's third clip, Lily also says 1 but it displays 4 😂
I got wrong in all except sykkuno 😂 I got it all I dunno why maybe because Iwatch too much Syk live here 😂
I want to see it where you ask people about their own voices, that would be hilarious
27:13 Toast should have been given another point. He should have had 2, not 1.
I got all the sykunno ones right. And I don’t even watch him lol. I think his voice is so unique that it’s easier to tell
I enjoyed playing along choosing whats fake and real! Thanks for the this!
Also, I have hard time telling whats fake...Robots are taking over hahaha
It's not fair when they all sounded like it got recorded by a microwave
20:25 young Sykkuno lol
I struggled on this so hard!! I’m glad Toast did too, I feel less stupid 😂
I have watched too much Syk and Rae lmao
Lol the only ones I got right was rae and sykkuno. 😂😂😂 I've never seen toast be that off in a video this was awesome we need another one
I can just imagine the "Sponsor Toast" intro.
"Heh. Spawn sword toast"
20:21 LMAO
I got everything wrong basically lol
This is lowkey terrifying it got this good with 30 minutes of data. Imagine if you input hours of data.
TIL I've watched enough Sykkuno to tell the difference
The fact I also knew all syk ones is impressive for myself lol !
Sounded like old school Donkey Kong Country music in the background for a bit...I like it.
I love these random gameshow videos 👍
I can tell all Sykkuno clearly I am a syk fan
This was so great! Would love to watch a part !!!!
Once I realized the AI leaves out end of sentence breaths that people have, it became easy to tell even for creators who I never watched. I think once it can learn little mannerisms like that, then it's all over
i got all peter (since we didn't know what the last one is about, i'll say i at least chose an audio that is fake), i got the last two of john and sydney, and got celine
i failed to recognize rae, yvonne and sykkuno, and i have probably seen more of them than john, syd and celine lol
the cooking mama music for johns, nice touch
Didn’t Scarra just win pre-tiebreaker? I don’t think Lilly got that last point.
Toast never gave himself a point for getting the first question right
this is a big brain play by toast......corpse probably
26:48 reveals how tirrred Toasts's editors are. No probs bro. We all saw Lily's 1. I hope Toast's dungeon is treating you well
I hear literally no difference between the fake and the real ones, it's absolutely crazy
the AI versions are really distinct, it's the night and day difference between the scripted expected way a person would say a line and the real way that a person produces speech. the ai samples don't have the same range of volume or pitch