Microsoft unveils VASA, an AI that can generate hyper-realistic talking face videos with precise lip-audio sync, lifelike facial behavior, and naturalistic head movements in real-time, from just a single portrait and voice clip.
>We introduce VASA, a framework for generating lifelike talking faces of virtual charactors with appealing visual affective skills (VAS), given a single static image and a speech audio clip. Our premiere model, VASA-1, is capable of not only producing lip movements that are exquisitely synchronized with the audio, but also capturing a large spectrum of facial nuances and natural head motions that contribute to the perception of authenticity and liveliness. The core innovations include a holistic facial dynamics and head movement generation model that works in a face latent space, and the development of such an expressive and disentangled face latent space using videos. Through extensive experiments including evaluation on a set of new metrics, we show that our method significantly outperforms previous methods along various dimensions comprehensively. Our method not only delivers high video quality with realistic facial and head dynamics but also supports the online generation of 512x512 videos at up to 40 FPS with negligible starting latency. It paves the way for real-time engagements with lifelike avatars that emulate human conversational behaviors.
 Thalidomide Vintage Ad Shirt $22.14 |
 |
Thalidomide Vintage Ad Shirt $22.14 |
>hyper-realistic
Paper written by:
>Sicheng Xu*, Guojun Chen*, Yu-Xiao Guo*, Jiaolong Yang*‡,
>Chong Li, Zhenyu Zang, Yizhong Zhang, Xin Tong, Baining Guo
yea, I guess it's a cultural difference
The hair physics are off.
That too the hair physics are bad
I wonder if this could be applied to games for things like applying realistic lip-sync to characters from voice-over audio.
I like how the number of teeth and their location keep changing if you keep looking at them.
Not to mention that, considering that it tries to mimic standard teeth, if you're a crooked teeth chad, anyone who has ever met you will immediately notice that a video with your face is fake even before they notice the poor coherency.
give expanding teeth gf
CGI/Motion did it better
Movement still looks robotic as shit.
Look like as if it's using one of those movement stabilization filters, that locks into a face and makes everything around it blurry. Also looks like shit
so can we assume glowies have had this tech for at least a decade?
You can watch the incremental research papers in real time bro. This is tech that is verifiably being developed right now.
Unnatural sliding motions
the motion is the same as how vtubers move
inhuman body language, but it is a start
Canadian man did it in the year 2001 but it didnt catch on and isnt as accurate as what microsoft did
Also, VASA is a city in Finland
>see this start shitting bricks
>see this no longer care
>b-but it'll improve
yeah yeah call me then homie
Don't care
Really it’s a huge relief that this AI craze popped only after boomers retired.
Now the ones in senior positions are Gen X, who are way more sober than the boomers who caused the dot com bubble.
so i can make a video of emma watson talking about how badly she wants to slurp my dick?
No, you can't. Only glowies wanting to manufacture evidence of crimes will be allowed full access to this AI. Everyone else gets the pozzed version because "we need to protect wahmen from being le heckin exploited!"
>Hyper realistic face
Maybe for an alien race with shifting morphing hair and beard hairs lol
The head movements aren’t even that right and the images don’t look realistic enough to be the real thing it needs to be in real time
Let me know when the bukkake DLC drops.
imagine the scams, sirs
none of these look convincing at all unless you're a moronic boomer
with that said, I'm expecting congress to shit their pants over it
They're not convincing if you look for it.
But they're good enough for disinfo and scams.
you say it is over as if i would ever be a woman to care about my job of "a pretty face that talks" being an artifact of the past. I will never be a woman thus i have none of the female problems.
btw what jobs will females now take given that they are losing all fronts: a troony takes their beauty pageant awards, an ai takes their girlfriend job, another ai takes their job... i mean we have like 1000 threads about it being over for programmers ( male ) but if i was a woman i would like actually consider the rope.
Ted was right.
>yfw the only way to avoid AI shananigans is to actively use various slurs (that AI is not allowed to use)
save democracy
say n*****
looks barely better than a mobile app
>hyper-realistic
>literally just photo morphing
Aww sweet, it's 2003 all over again!
Unironically what is the use case for this if not celebrity porn deep fakes and deep fakes of politicians saying things they didn't say?
Automatically animating codec calls in metal gear solid fangame
>capture dissident
>fake a video of him saying he's a pedophile and was actually running a child sex trafficking ring all along
>dissident is wholly discredited and his imminent death will go unreported by even the most rebellious of independent news sources
considering they’re all chinese it’s probably the real reason
So it's all shit got it.
Burn this tech with fire.
If they were gonna do that, all AI research would've been blackbagged in the crib. People know. People are gonna get all media evidence to no longer be evidence unless someone creates some sort of untamperable "this is real" watermark to confirm its really real and t
crap
thats how they're gonna do it
they're just gonna deepfake you and stamp it with the backdoored govt untampered watermark
frick
WE KNOW THIS SHIT EXISTS
WHY WOULD WE BELIEVE ANYTHING NOT IN PERSON?
>we
"we" is like 5% of the population. normies will eat any shit fed to them. but the feds don't even need this tech, they already just label any dissident a "rapist" and normoids believe it.
this tool is for corpo fricks, not feds. if you thought ads were annoying before, just wait until this shit pops up with realtime AI Black folk talking to you in personalized ads with the AI having access to your personality profile, credit card history, and internet history, all to persuade you to open up your wallet for the latest consumer garbage.
>Send out mass meeting invites to conglomerate employees as a 1:1 with CEO
>Stream prerendered video of CEO explaining an urgent need for their credentials to the system or company will lose millions
>1/5,000 falls for it
>Compromise system
>???
>Profit!!!
JUST BECAUSE YOU CAN DOES NOT MEAN THAT YOU SHOULD
This
Genocide all techbros (or techBlack folk, it means the same thing) by law
why are they all so smug looking
breasts or GTFO
>it's over
It's only just begun.
Accelerate.
but why?
Sweet more tech that only makes life worse for the plebs
Here for behind get ready for scam in your name, obviously fake news, and trans using this to promote interracial relationship
frick bros
we're gonna have third-world ractors in racting booths acting out illustrated primers for our children within the decade
diamond cities when
time to invest in some high power IR LEDs and face masks
What possible use case is there for something like this? Other than to get Taylor Swift image to talk to you dirty so that media can write how horrible AI is and how it needs to be banned yesterday.