Current Events > see that thing on AI creating videos from pictures/portraits?

(+) Yeah! (+)

https://www.vice.com/en_us/article/qv7zkw/create-fake-videos-of-faces-samsung-ai-labs-algorithm

Researchers at the Samsung AI Center in Moscow developed a way to create "living portraits" from a very small datasetas few as one photograph, in some of their models.

The paper, "Few-Shot Adversarial Learning of Realistic Neural Talking Head Models," was published on the preprint server arXiv on Monday.

The researchers call this few- and one-shot learning, where the model can be trained using just one image to create a convincing, animated portrait. With a few more shotsas few as eight or 32 photographsthe realism improves even more.

Because they only need one source image, the researchers were able to animate paintings and famous portraits, with eerie results. Fyodor Dostoevskywho died well before motion picture cameras became commercially availablemoves and talks in black and white. The Mona Lisa silently moves her mouth and eyes, a slight smile on her face. Salvador Dali rants on, mustache twitching.

These "photorealistic talking head models" are created using convolutional neural networks: They trained the algorithm on a large dataset of talking head videos with a wide variety of appearances. In this case, they used the publicly available VoxCeleb databases containing more than 7,000 images of celebrities from YouTube videos.

This trains the program to identify what they call "landmark" features of the faces: eyes, mouth shapes, the length and shape of a nose bridge.

This, in a way, is a leap beyond what even deepfakes and other algorithms using generative adversarial networks can accomplish. Instead of teaching the algorithm to paste one face onto another using a catalogue of expressions from one person, they use the facial features that are common across most humans to then puppeteer a new face.

The researchers write in the paper that they recognize the applications for realistic face-avatars in video conferencing, gaming, and special effectsbut the uncanny valley often holds us back from fully embracing widespread use of face-avatars of real people. They hope that this work changes that, with its low source requirements and "perfect" realism.

---

bump
---

jfc
---

yeah, that's p crazy
---

I can see this being really cool for history and whatnot. But I can also see it tremendously fucking with someone that has lost a loved one and has been clinging onto a photo of them.
---

I never thought I'd live to see the Mona Lisa speak.

yea, the Mona Lisa one is kinda bugging me out
---

boxington posted...

yea, the Mona Lisa one is kinda bugging me out

Same, they are all impressive, but that one in particular is pretty hypnotic
---

boxington posted...

bump

---

Skynet getting closer to becoming self-aware.
---

Is it just me, or did they make Salvador Dali look like a cross between Mike Patton and Orlando Bloom?
---

Clearly, a Mt Rushmore quartet is in order.

On_The_Edge posted...

jfc

---

Yo Mona Lisa, could I get a date on Friday?
And if you're busy, I wouldn't mind taking Saturday
---

Can't wait for a bot to start doing this in selfie topics gyarados
---

^ that would be pretty cool

I wonder if something like this would be more refined, if it'll help with animated shows/features
---

Imaging when they'll be able to make entire films using just still photos.
---

another industry in danger of being automated
---

one more
---

I mean, 20 posts in and nobody has mentioned what we're all thinking? Fine, I'll say it, Winston Churchill porn, here we come! :D
---

MabusIncarnate posted...

I can see this being really cool for history and whatnot. But I can also see it tremendously fucking with someone that has lost a loved one and has been clinging onto a photo of them.

There's research going on where you can take a video clip from anyone and if the angles are right record your own dialogue and make it look like they said it.

We're less than a decade away from an era where everything (video, pictures, stories, etc) can be synthetically generated to make anyone we want say anything we want or be anywhere we want in any picture or video with nearly no way to tell it was spoofed.

IF you think fake news is bad NOW, just wait.
---

This one, please:

no pls
---

This is amazing. Reminds me of those paintings in Harry Potter that could talk to you and interact with you.
---

One step closer to fucking the Mona Lisa.
---

Mist_Turnips posted...

One step closer to fucking the Mona Lisa.

You do know the Mona Lisa was a barely feminized daVinci self portrait, right?
Leonardo may well have invented rule 63.
https://abcnews.go.com/GMA/leonardo-da-vincis-mona-lisa-self-portrait/story?id=9662394

s0nicfan posted...

MabusIncarnate posted...

I can see this being really cool for history and whatnot. But I can also see it tremendously fucking with someone that has lost a loved one and has been clinging onto a photo of them.

There's research going on where you can take a video clip from anyone and if the angles are right record your own dialogue and make it look like they said it.

We're less than a decade away from an era where everything (video, pictures, stories, etc) can be synthetically generated to make anyone we want say anything we want or be anywhere we want in any picture or video with nearly no way to tell it was spoofed.

IF you think fake news is bad NOW, just wait.

Yep.
---

eggcorn posted...

Yep.

Luckily, I look pretty much like every other overweight scottish-american guy.
...or a hobbit, which is essentially the same thing.

One step closer to hogwarts
---

Questionmarktarius posted...

Mist_Turnips posted...

One step closer to fucking the Mona Lisa.

You do know the Mona Lisa was a barely feminized daVinci self portrait, right?
Leonardo may well have invented rule 63.
https://abcnews.go.com/GMA/leonardo-da-vincis-mona-lisa-self-portrait/story?id=9662394

Good
---

pretty creepy tbh
---

porn is going to be lit.
---

this brings porn to a new level

*imagines what this means for waifus*


---