OpenAI recently added a new Outpainting feature to their DALL-E artificial intelligence model. With its help, the neural network can expand the boundaries of paintings and any other images, drawing something on a specific user request.
For demonstration, the developers used the painting by Johann Vermeer “Girl with a Pearl Earring”. DALL-E, with the help of human prompts, “imagined” that he was behind the scenes of the portrait, and gradually completed it.
Note that even from the limited information provided by the canvas itself, the neural network did the job, mimicking the author’s style – mimicking the shadows and highlights of the original.
DALL-E users have appreciated Outpainting and have already published many really interesting extended versions of famous paintings and illustrations. Here is some of them.
1. “Napoleon at the St. Bernard Pass”, Jacques Louis David
It is difficult to distinguish the work of a neural network – a very accurately supplemented original of a famous painting.
2. An old advertising poster for the Quaker Oats brand
The author obviously mocked with tips, but it turned out funny.
3. “American Gothic”, Grant Wood
And here a couple of farmers turned out to be a real masterpiece.
4. “Dogs play poker” by Cassius Coolidge
The original picture is good in itself.
But with DALL-E, the picture literally sparkled with new colors.
5. Midnighters, Edward Hopper
Here, the neural network accurately determined the style and retained it while expanding the image borders.
6. The Beatles at the Abbey Road crossing
The famous frame and cover of the band’s self-titled album.
And this is how it could look when shooting with wide-angle optics.
7. “Mona Lisa”, Leonardo da Vinci
The original is known to everyone, but you hardly saw the version with a UFO, fire and some kind of monster.
8. Paramount Screensaver
The same mountains, but the view is farther away.
9. Fight Club Finale
Where Is My Mind?
10. Meme with a girl on the background of a fire
And finally, an extended version of the legendary meme, which confirms the guesswork about the causes of the fire.
Read also 🧐