How do you figure? With exception of "The Girl with a Pearl Earring" - out of the other six paintings, the Mona Lisa is the one we have no real, definitive evidence to show what she truly looked like. While it may be commonly believed that Lisa Gherardini was the muse for Mona Lisa, we have little/no evidence of what she looked like either - other than the Mona Lisa painting, which can't prove itself. As for the others: -* Frida Kahlo: there are numerous photographs and self portraits of Frida Kahlo throughout her life. -* "American Gothic": based Nan Woods (artist's sister) and Dr. Byron McKeeby (local dentist) and you can find a black and white photo of them standing in the same pose. -* "The Lady with Ermine": based on Cecilia Gallerani. There are several other paintings available of the same woman by other artists. -* "The Nightwatch" based on Frans Banninck Cocq and there are a couple other portraits of him also. -* "Birth of Venus": based on Simonetta Cattaneo Vespucci and also has multiple pieces of art depicting her likeness.
Please do the faces of real historical figures, like, Alexander the Great, Genghis Khan, Richard the Lionheart, Tutankhamun, Nefertiti, Cleopatra, etc.. Thank you!
Based on the one survived bust statue Nefertiti certainly was at least by contemporary standards a looker. ; ) Cleopatra then was like a crook nosed witch, but apparently a charming person and socially very clever. Plus a ruthless murderer and preferring incest to losing power...
People are being so critical. I don’t think they truly understand the gravity of what’s going on here. You are giving a machine a painting, and it is doing it’s best to turn it into a real life person, and then create a short video from that image. If people don’t understand how insane that is, they can’t be helped.
It's totally insane. Imagine you can have a realistically looking avatar using AI rendered portrait based on your favorite painting. Think for example about auto portrait of Albrecht Dürer... wow.
There's really nothing so remarkable about it, considering that 3d animation has existed since the early 1990's. In most of the above examples, the computer models resemble the original paintings very little.
Countalma And I’m sure the paintings resembled the actual models very little. This isn’t just 3D animation. It’s a robot mind analyzing the painting and using machine learning to turn it into a photo of a human being. Context matters. Jumping onto a box and doing a backflip off of it isn’t very remarkable either. People in gymnastics do it every day. But there’s a reason why Boston Dynamic’s robot doing it is pretty damn insane to watch.
I know right! If people actually knew the full math & full steps involved to do even part of this ..and actually maybe wrote just a part of it they would def approach their wording differently to this. Always the ones that know the real work involved in something that have these types of opinions haha :P .... Can even discuss the same points but approach it with more appreciation of whats already happening and what would have be involved to do it in the way the've suggested
Very sad to hear that historical footage which is definitely not supposed to be copyrightable anymore is causing your videos to suffer. Copyright law needs to be majorly curtailed.
We need something like Moore's Law for copyright. Since information transfer is getting 2x faster every 12-18 months, copyright terms should be halved every 12-18 months.
If you took any human that has lived in the past thousand years and took them a few hundred years into their future, they would be absolutely gobsmacked.
@@AtlasReburdened It's likely that humans 200 years from now will look back on us and be gobsmacked. The general consensus is that humans reached their peak around the late 80's. We are not only stagnating but actually going backwards as a species.
@@BartechTV The general consensus amongst whom, and by what metric? Certainly not the globally aggregated statistics regarding war, famine, death by disease, violent crime, and life expectancy.
@@miguelcrtz " people in real life dont look like the paintings, they are an approximation " Go look at the real Frida. Real photos exist of her. She looks nothing like the AI animation. Frida was a Latin American woman. Nonwhite. The AI animation looks like some white fashion model from France.
It's interesting to see the aesthetic effect of the training dataset which I assume involved contemporary photography with models that had modern style makeup and fit a modern sensibility of beauty. I wonder how these recreations would differ if they were trained on datasets from different decades of portrait photographs. How would the aesthetics vary, I wonder?
I had the same impression. They all have the same modern day make-up. Also the facial expressions seem very Western if not to say American (although the guy who did this is Russian, maybe his data sets were mostly American?). Also how do you get from the Mona Lisa painting with light brown eyes to an AI rendering with dark blue/grey eyes? This seems like crap science, honestly.
Muglife is too much fake fx... I mean, there is another process behind... MugLife works with aí, but just make a common 3d face shape in to the scene.. then the app has some video templates. I made first order model working via collab, but it is another experiment, not comparable to this one...
That is ridiculous about the Berlin video. They should be praising you and asking you to work with them. I'm so grateful for your hard work and that I was able to watch it. It was amazing!
LMAO. To be fair, it would make sense for them to be pretty as they were models of famous painters, so obviously those painters would choose beautiful women.
@@rafaelavega4923 They don't have to be fashion models of today to be "beautiful" or pretty. A lot of fashion models of today do look like that without makeup. The images we saw in this video, they look quite normal or plain. It only seems "pretty" because the skin appears to be flawless, but that's the result of A.I smoothing the image. While standards were different, there was always a distinction of at least plain and ugly people. What made people seem more attractive was status. That's why fat people in certain eras were considered beautiful but even then, they would be at worst, plain looking, but still definitely have an "appealing" face to them and a lot of times an appealing face meant their skin looked fair and healthy because "peasants" often worked under the sun and were malnourished in comparison, so they didn't have as healthy skin.
@Custom Chevys What? Are you serious? Of course she is Italian!, but Italians aren't Scandinavians, look that exaggerated white skin the A.I puts on her. That's what I mean with the A.I. database is biased
Hello Denis, I think your work is amazing and, simply put, of great historical value, which should supersede copyright claims. That video of Berlin was jaw dropping, I went and dragged my wife before the screen, a wide open window into the past. It put peace into my soul! I immediately downloaded it - for solely private use and pleasure only, of course. Love your accent, too. All the best!
There is a photo of the real-life American Gothic couple posed together in front of the painting. They were actually the artist Grant Wood’s Dentist and his daughter. Believe it or not, the painting probably flattered them.
It's a pity the networks didn't try to attribute the realistic features to the aesthetic of the portraits. Almost none of the subjects had make-up, yet the system autobooked it, making them seem nothing like the portraits. Beauty was vastly different when these artworks were made and differs from today standards of instagrammable glam.
Real question is how difficult for it to be for the algorithm NOT to use photos with heavy makeup not to mention if we have enough pictures of historically accurate photos for the algorithm to use.
@@reactions5783 that was my first pet peeve. Lol. I also didn't understand why the AI chose to make her hair brown instead of red with strawberry blonde highlights.
Totally agree, and don't forget that this is a camera obscura painting which means the portrait is at least 95% accurate to the model. So in a way the technology failed to create a face from a photo!
It is pleasure as well as honor for me to get to be a witness of such a great work and see all artistic emotions put in it. Kudos to you Denis! I hope in the future we'll meet and have a coffee together.
"Torture you with my accent" Funny Denis. I had an Indian prof for a Mechanics of Materials class. Funny guy, fairly thick accent. One day he informed the class he was thinking about teaching English the next semester. Everyone laughed. He looked hurt and surprised and perfectly straight faced, asked "What's so funny?" Obviously everyone laughed again. He was a good professor, too. One of the few classes I actually enjoyed. Interesting video, I enjoyed it. With the Mona Lisa, as one example, the teeth were obviously a guess, if the transformation was done with no other data than that painting available. What I suspect they do in that case is, if there are no other portraits or pics available of any given subject, is probably just average the teeth of the similarly shaped faces in the database, which could be accurate, and could be total crap. The AI probably also adjusts the teeth somewhat based on the facial shape, again, maybe a good educated guess, maybe crap. In the case of the Mona Lisa model you can see some pretty flaky looking stuff going on in the animated portion, which makes me question even more how accurate the faces are...just sayin'. I will go do more research. If the person can be located using facial recognition on the web, obviously they just use that, and if a fair amount of data is available, the face should be pretty much photo realistic. Just some guesses on methods (algos) employed, but they seem reasonable for at least a starting stab at the method(s). I suspect all of these are far better than a police sketch, but still short of the accurate face, some more than others. I went at looked at some actual pics of Frida Kahlo, and it got some fine details correct and some fairly grossly wrong. That's the problem with this AI stuff. All I hear is how immensely 'smart' it is, yet it fails to go do to ONE obvious piece of the puzzle that instantly occurred to me, there. Why? I went and listened to some of the AI discussion from the Google AI that was recently presented, and some of it is fairly impressive, except that I'll bet I can go find writings that virtually exactly emulate what it 'said', word for word. I'm still not convinced that that discussion is just cut and pastes from a look up table of a conglomeration of the web. Some of it was downright flaky, meaning very stilted and non human like. I realize that humans aren't really the necessary model, but just my take. I get the impression that true sentience and most certainly true self awareness, or the claims of self examination are on the seriously far fetched side. It would be interesting to see some real faces where no other data is available vs the reality.
@@Mr_Uni it her hair red in the transformed image? It looked brown to me, but then again I am legit colorblind, so red and brown look very similar to me.
Interesting take on famous people. Btw, your English grammatically perfect which is quite a feat . Yes of course bc English is your second language, you have a very charming accent which is very understandable. Kudos to you!
my thoughts exactly. Why bother applying AI to a portrait to get what the subject looked like "for real" when photos of Frieda Kahlo are quite numerous and you don't need an AI to show you how she looked... That particular one is rather pointless. The other treatments are interesting, but having had the great good fortune to see Monna Lisa from very close 3 weeks ago (about 1 meter away !) , I still prefer her as Leonardo Da Vinci painted her, so fine and detailed.
First time I have actually thought to myself Lisa Gherardini (Mona Lisa) must have been a VERY beautiful woman, I have always wondered why she was chosen because she looks so plain in the painting.
It was done 18 years ago. Google "real face of jesus" to see the result. He probably looked like a regular Semitic person in the area at that time, not like a northern European.
@VonAllen POV you don't be stupid. all you have to do is look at any thrift store portrait or kitschy item or mass-marketed jesus item circa ~1950-2008 to see a blue-eyed, light brown hair, light skinned version of jesus that serves the comfort-narrative of him not being 'othered.' your hyperbole doesn't hold up here.
@@DanielBoonelight I misinterpreted the comment on my last comment so I deleted it. Yes, you’re very much correct! It’s a complete distortion for visual portrayal. It’s kind of like propaganda when you think about it Edit: Spelling mistake
I know you don't claim to reproduce realistic historic lookalikes here, but I'd like to see you try. While the technology is great, the results were too prettified for me and too conforming to 21st century beauty ideals. In the case of American Gothic, for example, the outcome actually takes something away from the realness and impact of the original. Too polished what you did here. Too ad-compatible. But great job.
Would be nice to have an estimation how accurate those predictions are. Maybe try to apply this NN's to paintings of living people so that we have a reference
2nd painting is Frida Kahlo and there are tons of photos of hers. It could be expected that her real photo would be in video, but for some reason there's none. Video is great, nonetheless:)
I just compared the Frida Kahlo one to photos of her. I think Denis needs to add some irregularities to these to make them more life-like. The real Frida has something the image really lacks. Maybe it just proves that it's impossible to capture the human spirit in AI.
Watching the video is similar to observe the process of using the time machine. An unbelievable and surreal filling is being given with this masterpiece of neural link techs.
To me the Mona Lisa, the Mexican Woman and the Rembrand Nightwatch man are perfect. The others look either to old or just different. Anyhow I appreciated all your video and work and I subscribed, too. Thank you !!
mexican woman??? lol. It's Frida Kahlo, and there are lots of her photos exist, everyone knows how she looked. So this is total bullshit in this video to recreate her face.
You're amazing? Absolutely love it!. Also liked your using the same portraits used in animal crossing, I won't be able to look at them the same way. Keep up the good work!
Bruh when Mona Lisa smiled with those buck teeth, I fell in love. Quarantine has hit me hard The expressions really made them feel alive. Especially the odd looking ones
It would be ideal to get photos from people from the painters region or the origin of the subject and get a match based on historical parameters. It seems that the AI is picking on features on the ideal of beauty and how it is overly make up in both the painting and the reference images.. maybe training it with a set of specific people it would feel like it's extracting features from the gene pool.
Очень интересная идея, историческая реконструкция будет очень реалистичной, зная некоторые особенности и тд. В этих «оживлениях» как раз не хватает контекста эпохи и специфики некоторых персонажей, исторической достоверности, это поправимо вполне.
WHAT?? Is this uploader Russian? I had no idea he was Russian, I didnt even think his accent sounded Russian, but if he can read Russian comments then I assume he must be! Very surprising but at the same time, like many other people I find it VERY insulting to see comments on "English Speaking" videos in other languages! As just like in real life if a foreigner walks past speaking in a non-English accent & then for example continues to laugh its one of the most offensive things possible as we dont know if they could be talking about or insulting us, & its the same with your comment. Maybe u could "translate" it before posting next time?
@@VirusOfCyrus Why russians speaking russian is 'very insulting' to you? Imo op's surname and accent are as russian as it can get. Why would ommentor he translate it to english, the language he probably doesn't even speak? And his message was targeted towards op, not to some specific en speaking commentor who wouldn't understand.
To all the people who wonder why all the reconstructions look so beautiful: the algo's where probably trained with thousands of random faces. That means that if an algo reconstructs any face from a painting, when it has to make a choice, it the takes an average of all those faces. So the algo probably reconstructs the face to look more average (as in standard/mean/median, I dont mean 'plain'). Humans tend to find average faces to be more attractive. That is probably why all those recontructed faces look so beautiful. So it isn't like the AI's are conforming to beauty standards or anything (the painters probably did though).
omg they are so lifelike that I got overwhelmed by emotion and started crying, that kind of stuff really gets to me for some reason. Guess that's why I went to arts as my career. you should try 7 different portraits of Jesus next time! I'm really really curious what it would look like. Also, I don't mean to be nitckpicky, but for some portraits, I felt like the resulting faces were quite americanized and should've looked more European or Slav-ish. But I really don't know what databanks you use to train your AI so maybe I'm wrong. But this makes me think... it would be quite hard to make historically accurate faces, since all the high quality images on the internet are quite recent. And I guess finding enough material to fit historic-type portraits (un-photoshopped, no makeup, etc.) would be quite labor-intensive. I wonder what you think?
That is quite interesting. It is like you put the breath of life back into these long dead people and they were resurrected. I hope you do a couple more videos like this. These types of videos are quite mesmerizing.
These are awesome, but in all honesty: AI needs to come along a bit more before the creepy factor is gone for me. You do great work! I just think these are limited by the technology of today.
@@rosepetunia1829 Look at a photo of the real Frida Kahlo. She looks nothing like this AI animation. The AI animation looks like some white fashion model from France. Frida was nonwhite. Frida was a Latin American woman.
@@jotashock " Frida wasn't 100% Latina." That doesn't change a single thing I wrote, dummy. Like I wrote, LOOK AT A PHOTO OF THE REAL FRIDA KAHLO! Shes looks NOTHING like this AI animation. Frida was not a white woman. The AI model looks like a white woman. End of discussion.
Rose Petunia no way, there were so many inconsistencies, this was really fun to watch but in terms of how accurate it was, it was not very correct for most of them. Example would be Mona Lisa, her mouth was completely different shape and her nose around the tip is a little different, just slightly. Her nose looks more like a perfect nose, like the ones you see models have, but in the original her nose looks more defined and has a stronger shape that looks like it protrudes out more. As well as the shape around her brow, all these add up to make her look different, her eyebrows are also to low considering the shape of her eyebrow area bone
Part of it may be due to the selection of photos that were used as training data for the AI. Also, since the facial proportions and features in the paintings are not 100% accurate to real-life faces, it could be that the algorithm pulled each result toward the average/"ideal" face in the process of "correcting" them.
Now you can generate portraits here online:
neural.love/portraits
Have fun ✨
ratio
Hello Borat.
The Mona Lisa was the best one. She looked most like a real person and most like the painting it is based upon.
How do you figure? With exception of "The Girl with a Pearl Earring" - out of the other six paintings, the Mona Lisa is the one we have no real, definitive evidence to show what she truly looked like. While it may be commonly believed that Lisa Gherardini was the muse for Mona Lisa, we have little/no evidence of what she looked like either - other than the Mona Lisa painting, which can't prove itself.
As for the others:
-* Frida Kahlo: there are numerous photographs and self portraits of Frida Kahlo throughout her life.
-* "American Gothic": based Nan Woods (artist's sister) and Dr. Byron McKeeby (local dentist) and you can find a black and white photo of them standing in the same pose.
-* "The Lady with Ermine": based on Cecilia Gallerani. There are several other paintings available of the same woman by other artists.
-* "The Nightwatch" based on Frans Banninck Cocq and there are a couple other portraits of him also.
-* "Birth of Venus": based on Simonetta Cattaneo Vespucci and also has multiple pieces of art depicting her likeness.
@@Witchofthehill PhotoshopSurgeon did Monalisa and Cleopatra in real life, it looked better imo
She's hot as hell too lol. Goodness.
I know a girl who looks exactly like the Mona Lisa, and she does not look at all like this AI recreation.
@@guatagel2454 He at least got the eyes right.
Awesome video! The free AI portraits from kahma are stunningly realistic as well.
Don't tell my wife, but I think I just fell in love with Mona Lisa.
Steve M ok steve
@alphadawn2015 lennon wtf ewwwwahhh
Haha! Same here.
venus
Dude, she's a knock-out!
Wow the Mona Lisa looks just like the painting! I also tried kahma and was shocked by the results.
Cool concept! I also had fun recreating photos of myself and family at kahma
Please do the faces of real historical figures, like, Alexander the Great, Genghis Khan, Richard the Lionheart, Tutankhamun, Nefertiti, Cleopatra, etc.. Thank you!
Based on the one survived bust statue Nefertiti certainly was at least by contemporary standards a looker. ; )
Cleopatra then was like a crook nosed witch, but apparently a charming person and socially very clever. Plus a ruthless murderer and preferring incest to losing power...
Cleopatra 'real' faces were depicted in the coins,
I want Napoleon bonaparte and Frederick the great!!!!!
This!!!
Such a video exists already by channel MysteryScoops l :) !!
People are being so critical. I don’t think they truly understand the gravity of what’s going on here.
You are giving a machine a painting, and it is doing it’s best to turn it into a real life person, and then create a short video from that image. If people don’t understand how insane that is, they can’t be helped.
It's totally insane. Imagine you can have a realistically looking avatar using AI rendered portrait based on your favorite painting. Think for example about auto portrait of Albrecht Dürer... wow.
There's really nothing so remarkable about it, considering that 3d animation has existed since the early 1990's. In most of the above examples, the computer models resemble the original paintings very little.
Countalma And I’m sure the paintings resembled the actual models very little.
This isn’t just 3D animation. It’s a robot mind analyzing the painting and using machine learning to turn it into a photo of a human being.
Context matters. Jumping onto a box and doing a backflip off of it isn’t very remarkable either. People in gymnastics do it every day. But there’s a reason why Boston Dynamic’s robot doing it is pretty damn insane to watch.
Thank you lol
I know right! If people actually knew the full math & full steps involved to do even part of this ..and actually maybe wrote just a part of it they would def approach their wording differently to this. Always the ones that know the real work involved in something that have these types of opinions haha :P .... Can even discuss the same points but approach it with more appreciation of whats already happening and what would have be involved to do it in the way the've suggested
Now try to do Picasso’s paintings 😂
nightmare fuel
That would result in some serious silent hill stuff
My brain hurts
Jackson pollock might be interseting
For god's sake, don't XD
These are amazing! I like the results from kahma just as much, maybe more since theyre in vivid 8k.
its amazing to know that old photos can be restored to a new look!!! thanks for posting!!!
Very sad to hear that historical footage which is definitely not supposed to be copyrightable anymore is causing your videos to suffer. Copyright law needs to be majorly curtailed.
shouldn't they be in the public domain?
We need something like Moore's Law for copyright. Since information transfer is getting 2x faster every 12-18 months, copyright terms should be halved every 12-18 months.
They're a bunch of crying academic types.
Blame Disney
Agreed! That is an amazing restoration to a deterating important historical film, and gave us a glimpse into our past. Such a shame.
This channel is actually run by a neural network.
It is :-)
Ha-ha, BTW can you solve a few CAPTCHAs for me, please?
@@DenisShiryaev just do them slower Mr. AI
If all the painters were still alive, they would say *Holy crap, what kind of magic do you use*
If all the painters were still alive, i would say Holy crap, what kind of magic do they use =D
If you took any human that has lived in the past thousand years and took them a few hundred years into their future, they would be absolutely gobsmacked.
@@AtlasReburdened It's likely that humans 200 years from now will look back on us and be gobsmacked. The general consensus is that humans reached their peak around the late 80's. We are not only stagnating but actually going backwards as a species.
@@BartechTV The general consensus amongst whom, and by what metric? Certainly not the globally aggregated statistics regarding war, famine, death by disease, violent crime, and life expectancy.
If all the painters were still alive, I would say *Holy crap! Why all of you are still alive?*
Thank you for transforming these paintings. I was enchanted.
I’m in love with the Mona Lisa girl , Amazing effects!
This is cute but am I the only one who thinks the moving ones look nothing like the originals?
I also expected that the original would be kept. It is disappointing that only some faces were simply taken for the Software.
people in real life dont look like the paintings, they are an approximation
@@miguelcrtz " people in real life dont look like the paintings, they are an approximation "
Go look at the real Frida. Real photos exist of her. She looks nothing like the AI animation. Frida was a Latin American woman. Nonwhite. The AI animation looks like some white fashion model from France.
@@miguelcrtz 1:37 the virtual one is much thinner, and also have shorter chin. At least that shouldn't have happened
I'm guessing that he was trying to glamorize them, and he got the lines wrong. I don't even recognize them a characters.
The mona lisa smiling is killing me. So cute.
Wow. I would ask the mona lisa out for sure
And he would probably say no for sure! 8-)
Thom Gillespie Harsh (but probably true). 😁
@@MaggieTheCat01 I was making a 'gender' comment since some think The Mona Lisa was not a female model. I actually love the work.
Thom Gillespie Mona Lisa is trans!?
Thom Gillespie Oh, I thought you simply did a typo. 👍
Watching The Night Watch come together was amazing to see. Keep up the amazing work!
I love this video concept! Would enjoy a series! Thank you for sharing! ❤️
It's interesting to see the aesthetic effect of the training dataset which I assume involved contemporary photography with models that had modern style makeup and fit a modern sensibility of beauty. I wonder how these recreations would differ if they were trained on datasets from different decades of portrait photographs. How would the aesthetics vary, I wonder?
I had the same impression. They all have the same modern day make-up. Also the facial expressions seem very Western if not to say American (although the guy who did this is Russian, maybe his data sets were mostly American?). Also how do you get from the Mona Lisa painting with light brown eyes to an AI rendering with dark blue/grey eyes? This seems like crap science, honestly.
oooooooo, imagine an 80s style Girl with a Pearl Earring lmao
I straight up fell in love with the Mona Lisa one.
If I send you a painting, will you animate it like the others?
Download the mug life app and you can do it yourself!
Muglife is too much fake fx... I mean, there is another process behind... MugLife works with aí, but just make a common 3d face shape in to the scene.. then the app has some video templates.
I made first order model working via collab, but it is another experiment, not comparable to this one...
@@GiorgioTedesco do you k ow any app that can do the same as this video?
she thicc tho
Your are great Dennis! Love your accent and vids. I love the feeling the we are seeing these people again after all those years
So the Girl with the Pearl Earing is basically Scarlett Johansson.
I think we all believe that
She actually have a movie based on that
… and The Nightwatch shows Gerard Depardieu …
Frida Carlo looks like a Paris fashion model
Yep, and actually male
Kahlo
and her AI version too!
Yes🙈😂
I love Frida Carlo shes so crazy i love her
That is ridiculous about the Berlin video. They should be praising you and asking you to work with them. I'm so grateful for your hard work and that I was able to watch it. It was amazing!
Why are these people all so pretty?
This AI is thirsty.
LMAO. To be fair, it would make sense for them to be pretty as they were models of famous painters, so obviously those painters would choose beautiful women.
Demon Seed was a movie about AI. The computer system ended up fertilising Julie Christie. 🤔
@Archock Encanto well said
@@khazms the standards of beauty were different back then so i doubt all these people looked like the fashion models of today
@@rafaelavega4923 They don't have to be fashion models of today to be "beautiful" or pretty. A lot of fashion models of today do look like that without makeup. The images we saw in this video, they look quite normal or plain. It only seems "pretty" because the skin appears to be flawless, but that's the result of A.I smoothing the image.
While standards were different, there was always a distinction of at least plain and ugly people. What made people seem more attractive was status. That's why fat people in certain eras were considered beautiful but even then, they would be at worst, plain looking, but still definitely have an "appealing" face to them and a lot of times an appealing face meant their skin looked fair and healthy because "peasants" often worked under the sun and were malnourished in comparison, so they didn't have as healthy skin.
Seems like the algo was only trained with pretty people. The simulation always looks more attractive than the drawing. 😂
I think you are right, and maybe why the farmer’s wife was enhanced and the face was truly not her own.
not really 'more' i think the AI got trained with current beauty standards which is why it deviated significantly with the artists' intent.
More like white people database
@Custom Chevys What? Are you serious? Of course she is Italian!, but Italians aren't Scandinavians, look that exaggerated white skin the A.I puts on her. That's what I mean with the A.I. database is biased
@@CamiloSoto Italians may have a very pale complection, as a significant part of the Portuguese. Pale with dark hair and eyes.
You’re amazing. I wish I could attend an entire gallery of your works.
Thank you so much 😀
Fascinating! Great work Denis .
What this algorithm really does is to make the ladies look a lot prettier, even if it doesn't stay 100% true to the originals.
Because it got trained with pretty instagram stars. If the algorythm thinks 100% of women are pretty thats how its going to generate them.
I think, the models for original paintings would not be as pretty as the paintings themselves.
My impressions are quite the opposite! Somehow the author of the video made them ugly and without true resemblance to the originals!
Hello Denis, I think your work is amazing and, simply put, of great historical value, which should supersede copyright claims. That video of Berlin was jaw dropping, I went and dragged my wife before the screen, a wide open window into the past. It put peace into my soul! I immediately downloaded it - for solely private use and pleasure only, of course.
Love your accent, too. All the best!
Is there anyway to get the Berlin Video please? I guess I missed it and really want to watch it.
alexander brand How long is a copyright?
@@Dietrichunder me too
NichtsNeues. ImWesten me too
Please put in a Google drive and share link!!
So this is what the paintings would look like if they were those in Harry Potter and could move? Wonderful what muggle technology can do!
This I like.
Close enough to harry potter
Muggles learned the magic! RIP VolDeMort
There is a photo of the real-life American Gothic couple posed together in front of the painting. They were actually the artist Grant Wood’s Dentist and his daughter. Believe it or not, the painting probably flattered them.
Great work Denis! Cheers from Canada.
It's a pity the networks didn't try to attribute the realistic features to the aesthetic of the portraits. Almost none of the subjects had make-up, yet the system autobooked it, making them seem nothing like the portraits. Beauty was vastly different when these artworks were made and differs from today standards of instagrammable glam.
Yes, the originals actually look more innocent.
Real question is how difficult for it to be for the algorithm NOT to use photos with heavy makeup not to mention if we have enough pictures of historically accurate photos for the algorithm to use.
I guess the network feeds of nowadays photos so... yeah makes sense.
That's why all the faces looked like movie stars or magazine models, not like real people.
Agreed, it's the flaw in the training source images, it would have to adjust for any portrait enhancements, lighting, and skin shades as well.
So the girl with a pearl earring was more like a Winona Ryder than a Scarlett Johansson ?....
Yea. I always imagine her to look a bit like Lady Gaga, but maybe less long face?
2:55 Most of these AI Paintings are unrealistic and they make European woman to Asian.
@@Larrypint They also changed the Venus hair at 3:00 from strawberry blonde to a dark brown. Why can't the AI simply keep the correct shade of hair?
@@reactions5783 that was my first pet peeve. Lol. I also didn't understand why the AI chose to make her hair brown instead of red with strawberry blonde highlights.
Pearl Earring girl looks NOTHING AT ALL like her painting. Lol
She looks like a Final Fantasy character 😁
Totally agree, and don't forget that this is a camera obscura painting which means the portrait is at least 95% accurate to the model. So in a way the technology failed to create a face from a photo!
I agree hahaha
I'm surprised the AI didn't make the girl look like Scarlett Johansson
Yep, I know someone who looks exactly like the girl with a pearl earring. She's way more attractive than the AI version.
It is pleasure as well as honor for me to get to be a witness of such a great work and see all artistic emotions put in it. Kudos to you Denis! I hope in the future we'll meet and have a coffee together.
Good stuff. Enjoyed. And your sense of humor was a +
7:19 Random guy from the internet - Denis Shiryaev himself.
Thats a dapper dude.
Лойс Денису за чувство юмора и скромность))
Попався
😂
still I got found a Girl looks alike 'mona lisa' so far... check it Instagram: @caro.tobon15
you're great, denis!
The part when they “wake up” is so freaky
They remind of Sims characters in the creation section
Okay so it's not just me. I see this potentially used in horror movies...paintings waking up and saying something....presumably..something scary.
"Torture you with my accent" Funny Denis.
I had an Indian prof for a Mechanics of Materials class. Funny guy, fairly thick accent. One day he informed the class he was thinking about teaching English the next semester. Everyone laughed. He looked hurt and surprised and perfectly straight faced, asked "What's so funny?" Obviously everyone laughed again. He was a good professor, too. One of the few classes I actually enjoyed.
Interesting video, I enjoyed it. With the Mona Lisa, as one example, the teeth were obviously a guess, if the transformation was done with no other data than that painting available. What I suspect they do in that case is, if there are no other portraits or pics available of any given subject, is probably just average the teeth of the similarly shaped faces in the database, which could be accurate, and could be total crap. The AI probably also adjusts the teeth somewhat based on the facial shape, again, maybe a good educated guess, maybe crap. In the case of the Mona Lisa model you can see some pretty flaky looking stuff going on in the animated portion, which makes me question even more how accurate the faces are...just sayin'. I will go do more research.
If the person can be located using facial recognition on the web, obviously they just use that, and if a fair amount of data is available, the face should be pretty much photo realistic. Just some guesses on methods (algos) employed, but they seem reasonable for at least a starting stab at the method(s). I suspect all of these are far better than a police sketch, but still short of the accurate face, some more than others. I went at looked at some actual pics of Frida Kahlo, and it got some fine details correct and some fairly grossly wrong. That's the problem with this AI stuff. All I hear is how immensely 'smart' it is, yet it fails to go do to ONE obvious piece of the puzzle that instantly occurred to me, there. Why?
I went and listened to some of the AI discussion from the Google AI that was recently presented, and some of it is fairly impressive, except that I'll bet I can go find writings that virtually exactly emulate what it 'said', word for word. I'm still not convinced that that discussion is just cut and pastes from a look up table of a conglomeration of the web. Some of it was downright flaky, meaning very stilted and non human like. I realize that humans aren't really the necessary model, but just my take. I get the impression that true sentience and most certainly true self awareness, or the claims of self examination are on the seriously far fetched side.
It would be interesting to see some real faces where no other data is available vs the reality.
The night watch guy is definitely my guy type!!!! Great job and FUNtastic!!!!!
You should do this to portraits of famous kings and queens. Good Job! It is awesome!
Would like to see too
This!
LOL, Venus ended up looking like a Kpop starlet.
Yeah, she just lools ugly. Way worse than the actual picture. Also why did the algorithm change her hair color and teint?
Captain BaseBallBat-Boy That’s exactly what I was saying. Isn’t her hair red in the painting?
@@Wolf_Larsen wtf she looks cute to me
@@Mr_Uni it her hair red in the transformed image? It looked brown to me, but then again I am legit colorblind, so red and brown look very similar to me.
Captain BaseBallBat-Boy it looked brown to me in the transformation.
I'm thoroughly creeped out by the AI's baring their teeth while shaking their heads. Wtf
The paintings are becoming e-girls
could say the same about your profile pic
@@bennett1426 Mona Lisa OnlyFans when lmao
Bro.. AMAZING! Simply Astonishing! MORE PLEASE!
Interesting take on famous people. Btw, your English grammatically perfect which is quite a feat . Yes of course bc English is your second language, you have a very charming accent which is very understandable. Kudos to you!
The Freda Kahlo one was a bit daft seeing as she was a real person. Just put her photo up.
So was like almost all of the others .
Your comment is so beside the point that I'm surprised you are even here watching this.
I'm just confused why make her white? Disappointing to see as a fellow Mexican :/
my thoughts exactly. Why bother applying AI to a portrait to get what the subject looked like "for real" when photos of Frieda Kahlo are quite numerous and you don't need an AI to show you how she looked... That particular one is rather pointless. The other treatments are interesting, but having had the great good fortune to see Monna Lisa from very close 3 weeks ago (about 1 meter away !) , I still prefer her as Leonardo Da Vinci painted her, so fine and detailed.
@@sylviecharlois1165 I think the creator wasn't well versed in her history.
The resemblance are impeccable!
I can’t tell if you mean that or if it’s sarcasm, sorry...
Looks nothing like originals.
@@jumpingblue1623 i mean.... the mona Lisa was the most realistic.
Mona Lisa looked like an actual Italian girl 😮 and Venus could easily look like a Greek girl 🤔
First time I have actually thought to myself Lisa Gherardini (Mona Lisa) must have been a VERY beautiful woman, I have always wondered why she was chosen because she looks so plain in the painting.
Alright it is time, please, show me Jesus
It's not #CesareBorgia
Most heavy musicians look like Jesus!
It was done 18 years ago. Google "real face of jesus" to see the result. He probably looked like a regular Semitic person in the area at that time, not like a northern European.
@VonAllen POV you don't be stupid. all you have to do is look at any thrift store portrait or kitschy item or mass-marketed jesus item circa ~1950-2008 to see a blue-eyed, light brown hair, light skinned version of jesus that serves the comfort-narrative of him not being 'othered.' your hyperbole doesn't hold up here.
@@DanielBoonelight I misinterpreted the comment on my last comment so I deleted it. Yes, you’re very much correct! It’s a complete distortion for visual portrayal. It’s kind of like propaganda when you think about it
Edit: Spelling mistake
I wonder why the AI reduces so much the size of the eyes for the "Girl with a Pearl Earring".
Beautiful work! Greetings from Colombia.
I know you don't claim to reproduce realistic historic lookalikes here, but I'd like to see you try. While the technology is great, the results were too prettified for me and too conforming to 21st century beauty ideals. In the case of American Gothic, for example, the outcome actually takes something away from the realness and impact of the original. Too polished what you did here. Too ad-compatible. But great job.
Would be nice to have an estimation how accurate those predictions are. Maybe try to apply this NN's to paintings of living people so that we have a reference
Pretty sure those paintings ARE accurate...
2nd painting is Frida Kahlo and there are tons of photos of hers. It could be expected that her real photo would be in video, but for some reason there's none. Video is great, nonetheless:)
I just compared the Frida Kahlo one to photos of her. I think Denis needs to add some irregularities to these to make them more life-like. The real Frida has something the image really lacks. Maybe it just proves that it's impossible to capture the human spirit in AI.
Watching the video is similar to observe the process of using the time machine.
An unbelievable and surreal filling is being given with this masterpiece of neural link techs.
What a great work, My friend. Congratulations!
This video is a great example of why AI is nowhere close to being ready. Neural networks are convoluted lookup tables.
Give it a few more years
Thats because neural networks arent actually AI, they are just learning algorythms.
the girl with a pearl earring turned out to be completely different from the original
To me the Mona Lisa, the Mexican Woman and the Rembrand Nightwatch man are perfect. The others look either to old or just different. Anyhow I appreciated all your video and work and I subscribed, too. Thank you !!
If that's what you saw, okay. I saw differences when looking left and right. It was good though, really good. But not perfect.
mexican woman??? lol. It's Frida Kahlo, and there are lots of her photos exist, everyone knows how she looked. So this is total bullshit in this video to recreate her face.
Wow! One day there will be movies that we’ll be watching and instead of actors will see personages from all these paintings, so amazing!
What a fun video and idea! I enjoy all of your videos immensely. Thank you!
Эх ещё бы бюсты и статуи. На Цезаря бы посмотреть...
Есть реконструкции антропологов очень хорошие, можно их оживить.
Поиграйте Assassin Creed. Ubisoft с сюжетом конечно мудрят, но с точки зрения воссоздания исторических персонажей и мест очень даже стараются
th-cam.com/video/M43iC46-XA0/w-d-xo.html
И исторические фигуры, такие как Генрих 8 и его жены и его дети и еще Мария Стюарт !
For "Girl with a Pearl Earring", Shelly Duvall (The Shining, 1980 era), would be an exact match.
You know, I had the same feeling when I first saw that too
Yeah I had the same thought!
Это просто обалденно!!! Очень здорово!! Хотим ещё!! Пожалуйста, пожалуйста, пожалуйста!!
You're amazing? Absolutely love it!. Also liked your using the same portraits used in animal crossing, I won't be able to look at them the same way. Keep up the good work!
Good craftswork, Denis!
Imagine if this technology was in museums... it would never be the same
would that the same girls ethnicity Antioquia - Colombia, appears look like actress 'Amparo grisales', IG: Caro.tobon15, so many Girls Paisa colombian
Bruh when Mona Lisa smiled with those buck teeth, I fell in love. Quarantine has hit me hard
The expressions really made them feel alive. Especially the odd looking ones
If I put my photo this algorithm would turn me into an attractive english man with a small nose.
Woah, that's awesome work in there! I really want more of it, like of Beethoven, Mozart, Chopin and Liszt!! And also famous egyptian rulers!
Fascinating! I could have watched this for hours!
оживи бурлаков на волге или казаков, пишущих письмо турецкому султану
TBicm
сразу видно наш человек) ага, сказал китаец) lmao.
На улицу выйди , там этого быдла полно везде , живых
Sorry, i don't speak enchantment table
@SUPERΔRTIFICIΔL ахахах, я по жизни
Nobody:
Russians: Sorry for my russian accent
It’s apparently a sensitive topic with Russians
almost... believe Russian-Qazaqatan ethics
It would be ideal to get photos from people from the painters region or the origin of the subject and get a match based on historical parameters.
It seems that the AI is picking on features on the ideal of beauty and how it is overly make up in both the painting and the reference images..
maybe training it with a set of specific people it would feel like it's extracting features from the gene pool.
This is definitely one of the coolest videos I’ve seen in a long time
Fantastic job! Really enjoyed and appreciate your work!
Очень интересная идея, историческая реконструкция будет очень реалистичной, зная некоторые особенности и тд. В этих «оживлениях» как раз не хватает контекста эпохи и специфики некоторых персонажей, исторической достоверности, это поправимо вполне.
WHAT?? Is this uploader Russian? I had no idea he was Russian, I didnt even think his accent sounded Russian, but if he can read Russian comments then I assume he must be! Very surprising but at the same time, like many other people I find it VERY insulting to see comments on "English Speaking" videos in other languages! As just like in real life if a foreigner walks past speaking in a non-English accent & then for example continues to laugh its one of the most offensive things possible as we dont know if they could be talking about or insulting us, & its the same with your comment. Maybe u could "translate" it before posting next time?
@@VirusOfCyrus Абсолютно согласен с тобой)
@@VirusOfCyrus Why russians speaking russian is 'very insulting' to you?
Imo op's surname and accent are as russian as it can get. Why would ommentor he translate it to english, the language he probably doesn't even speak? And his message was targeted towards op, not to some specific en speaking commentor who wouldn't understand.
To all the people who wonder why all the reconstructions look so beautiful: the algo's where probably trained with thousands of random faces. That means that if an algo reconstructs any face from a painting, when it has to make a choice, it the takes an average of all those faces. So the algo probably reconstructs the face to look more average (as in standard/mean/median, I dont mean 'plain'). Humans tend to find average faces to be more attractive. That is probably why all those recontructed faces look so beautiful. So it isn't like the AI's are conforming to beauty standards or anything (the painters probably did though).
U mean its an Agro algo
It may be trained with celebrity faces
Привет Денис! С удовольствием смотрю ваши творческие эксперименты. Вы не думали сделать русскоязычный канал?
+
Incredible. Great work. Maestro.
Денис, обалденные портреты, невероятно точное сходство с оригиналом!
omg they are so lifelike that I got overwhelmed by emotion and started crying, that kind of stuff really gets to me for some reason. Guess that's why I went to arts as my career.
you should try 7 different portraits of Jesus next time! I'm really really curious what it would look like.
Also, I don't mean to be nitckpicky, but for some portraits, I felt like the resulting faces were quite americanized and should've looked more European or Slav-ish. But I really don't know what databanks you use to train your AI so maybe I'm wrong. But this makes me think... it would be quite hard to make historically accurate faces, since all the high quality images on the internet are quite recent. And I guess finding enough material to fit historic-type portraits (un-photoshopped, no makeup, etc.) would be quite labor-intensive. I wonder what you think?
Bump.
Why would you cry about that?
Как летит время ... "деточка" подрос. Смотрите, ребята, не учите его плохому! Денис твоё создание впечатляет!!!
Mona Lisa: Hi
Me: How yo doin' ?
That is quite interesting. It is like you put the breath of life back into these long dead people and they were resurrected. I hope you do a couple more videos like this. These types of videos are quite mesmerizing.
Fascinating. This was well worth watching. I hope you do some more like it.
These are awesome, but in all honesty: AI needs to come along a bit more before the creepy factor is gone for me. You do great work! I just think these are limited by the technology of today.
Doesn't really look like them. Sorry. Dont see it. But I do see a value in this. Perhaps with adjustment.
How can you not see it? It looks 100% like them!
@@rosepetunia1829 Look at a photo of the real Frida Kahlo. She looks nothing like this AI animation. The AI animation looks like some white fashion model from France. Frida was nonwhite. Frida was a Latin American woman.
@@NomadFlow Frida wasn't 100% Latina. His mother was Mexican but his father was German.
@@jotashock " Frida wasn't 100% Latina."
That doesn't change a single thing I wrote, dummy. Like I wrote, LOOK AT A PHOTO OF THE REAL FRIDA KAHLO! Shes looks NOTHING like this AI animation. Frida was not a white woman. The AI model looks like a white woman. End of discussion.
Rose Petunia no way, there were so many inconsistencies, this was really fun to watch but in terms of how accurate it was, it was not very correct for most of them. Example would be Mona Lisa, her mouth was completely different shape and her nose around the tip is a little different, just slightly. Her nose looks more like a perfect nose, like the ones you see models have, but in the original her nose looks more defined and has a stronger shape that looks like it protrudes out more. As well as the shape around her brow, all these add up to make her look different, her eyebrows are also to low considering the shape of her eyebrow area bone
... is indistinguishable from magic.
Would not want to meet the DaVinci in a dark alleyway. Dude was ahead of his time
All beautiful and amazing, thank yoi!
Makes the ginger Venus a brunette.
All redheads: TRIGGERED.
Слушай очень круто, продолжай!
The woman in the animations are looking so much nicer than in the paintings. (Why?)
Based on modern people
Part of it may be due to the selection of photos that were used as training data for the AI. Also, since the facial proportions and features in the paintings are not 100% accurate to real-life faces, it could be that the algorithm pulled each result toward the average/"ideal" face in the process of "correcting" them.
because the AI has been programmed by horny men. this is why Mona Lisa looks and acts like porn star
@@cylmareall 😂
Crazy stuff. Keep it up! 👌