We crafted these weird and wonderful images using OpenAI’s Dalle-2

(Pocket-lint) – There are a few different AI technologies popping up lately that are able to create interesting and convincing images using artificial intelligence.  Google has one in the form of Imagen but the most well-known is OpenAI’s Dalle-2. This is a system that’s able to create realistic images and art based on descriptive terms input by the user. 

Dalle-2 is now open to the public meaning anyone can use it and for free (up to a certain number of uses). So we’ve been testing out the AI to see what images we could create. Keep scrolling to see what we’ve crafted.

We’ve included the terms we used to create the images, so you can see how close the AI got to meeting the request. 

POCKET-LINT VIDEO OF THE DAYOpenAI/Pocket-lint

Kid’s drawing

Dalle-2 is interesting because you can not only tell it what to create but also the style of art you’re after. So that’s just what we did: “Child’s crayon styling drawing of a monster on a wall like graffiti.”

OpenAI/Pocket-lint

Doughnut house

You can also let your imagination run away with you.

So if you’re hungry then just turn food into objects: “A house made entirely of ring doughnuts with a pink glaze and sprinkles.”

OpenAI/Pocket-lint

New Google Pixel watch

We thought it might be fun to try to get an idea of future smart devices and what an AI might think they looked like. Sadly it did a better job on the woman’s hair than the watch. 

“New Google Pixel watch with bright colours being worn by a stylish woman with curly hair.”

OpenAI/Pocket-lint

Batman’s new ride

You can make your requests as specific as you like right down to the era of something, the style or simply the subject matter.  “A new Batmobile with 1990s styling being admired by The Joker.”

OpenAI/Pocket-lint

Balloon monsters

What’s one of the least fearsome things you’re ever likely to see? A balloon monster?

We figured that might make a fun alternative to one of the other giant creatures that have historically attacked Japan.  “A giant green balloon animal attacks Tokoyo.”

OpenAI/Pocket-lint

Concept BMW

We love concept cars that various brands come up with. Designers and engineers are coming up with all sorts of future designs for cars.

But what can the AI think the future looks like? “A futuristic concept of an autonomous self-driving BMW on a quiet street.” 

OpenAI/Pocket-lint

Space monkey

What if we sent monkeys to Mars instead of people? What would that look like?

Well, according to Dalle-2 that would result in some fairly grumpy-looking monkeys. Understandably.  “High resolution photo of a monkey astronaut on his way to Mars.”

OpenAI/Pocket-lint

Germs

We recently wrote about Nikon’s Photomicrography Competition which features awesome photos of the tiniest things in our world.

We wanted to see what Dalle-2 could create.  “Award-winning photos of micro organisms.”

OpenAI/Pocket-lint

An exploding phone

Not one wants to see this happening to their smartphone, yet we’ve seen enough horror stories over the years.  “A mobile phone that’s overheated and exploded on a wooden table.”

OpenAI/Pocket-lint

Roses in bloom

Sometimes you’ve got to take time out to smell the roses, even if those roses aren’t real and were created by AI. 

Dashlane can keep your employees’ passwords safe By Pocket-lint Promotion . 23 August 2022

If you’re concerned about your organisation’s security, this is a no-brainer.

“A macro photograph of a rose bush in full bloom.”

OpenAI/Pocket-lint

Dogs in a submarine

Dalle-2 seems pretty good at creating cartoon imagery. In this instance an amusing vision of a pair of dogs having a great time under the sea.  “A cartoon dog inside a yellow submarine deep under the sea.”

OpenAI/Pocket-lint

Concept cars

Concept cars can often be a bit out there.

And that’s the fun of them because designers often have a bit more freedom in their creations. So why not let the AI have some fun too? “The weirdest concept car ever designed by man.”

OpenAI/Pocket-lint

Duck vs turtle

After seeing what Google Imagen could do we thought it would be interesting to replicate its results by using the same terms.

The results aren’t quite as impressive, but are still a lot of fun. “A chrome-plated duck with a golden beak arguing with an angry turtle in a forest.”

OpenAI/Pocket-lint

Robots in an art gallery

Another of Google’s examples saw some funky robots in a flooded gallery. What would Dalle-2 do with such a request?

“An art gallery displaying Monet paintings. The art gallery is flooded. Robots are going around the art gallery using paddle boards.”

OpenAI/Pocket-lint

Corgi in Times Square

Seeing dogs wearing hats and behaving like human beings is always fun in real life.

We love that the AI has chosen to embody that joy in one image while making the dog super grumpy in the other.  “A photo of a Corgi dog riding a bike in Times Square. It is wearing sunglasses and a beach hat.”

OpenAI/Pocket-lint

The moon made of cheese

We were frankly a bit disappointed with this one as we were hoping for something a bit more creative when we asked to see the moon made of cheese.

But perhaps you need to be a bit more specific in your requests to get the best results.  “The moon made of cheese.”

OpenAI/Pocket-lint

A space racoon

With this one, we actually got two quite different results, if slightly subtle ones. One racoon is looking inside from outside and one doing the opposite. 

“A photo of a raccoon wearing an astronaut helmet, looking out of the window at night.”

OpenAI/Pocket-lint

An exploding vase

We’re big fans of high-speed photography and slow-mo results. So we wondered what Dalle-2 would come up with if we asked it to create something similar.   “A glass vase full of flowers explodes in slow motion in front of a colourful wall and glass shatters all over the place.”

OpenAI/Pocket-lint

A cyberpunk styled car races away from police on horseback in Times Square

For some reason when we asked for this scene Dalle-2 gave the police officers space helmets.

Make of that what you will.  “A cyberpunk styled car races away from police on horseback in Times Square.”

OpenAI/Pocket-lint

Colourful computers

One of the highlights of this AI technology is the way it can create new artwork based on the style of a human artist. Even where that artist might not have painted such a picture before.  

“A painting of a 90s computer in the style of Claude Monet.”

OpenAI/Pocket-lint

Vader is angry

Admittedly Darth Vader looks a bit strange in a couple of these images, but the overall theme is still sort of what we were hoping for.  “Darth Vader angrily chastises a stormtrooper who is crying at his desk.”

OpenAI/Pocket-lint

Teddy time

There’s something amusing about Teddy Bears going on adventures.  These two furry creatures are having a great time on skate boards in public spaces.  “A posh teddy bear on a skateboard near the Arc de Triomphe.”

OpenAI/Pocket-lint

Cat monster

We’ve had fun experimenting with what Dalle-2 can create when it comes to cartoon artwork.

Here cats become sea monsters and frankly it’s quite unnerving.  “A cat seamonster destroying a fishing boat during a storm.”

OpenAI/Pocket-lint

Cybernetic humans in a pop art style

Even if the end results can be a touch hit and miss at times, you’ve got to admire how versatile the AI can be. How many humans could create so many different styles of artwork?

“A cybernetic human painted in pop art style.”

OpenAI/Pocket-lint

Aliens at the Whitehouse

It seems that artificial intelligence doesn’t have any idea what aliens might look like. We do find it interesting that they apparently can look like ghosts who travel by space jellyfish though.  “Alien visitors at the White house in the art style of Mars Attacks!”

Writing by Adrian Willings.