If you’ve been perusing the internet these past few months, you may have noticed an influx of AI-generated art, and there’s a good chance that some of it has come from Dall-E 2.
The world of technology is growing more advanced every day, and people are becoming better equipped to create technology that services specific needs and wants.
And one thing technology is getting better at is replicating human behaviour. Even behaviour some may consider strictly human, like creating art.
Dall-E 2 is a piece of software able to do just that, and it’s likely you’ve seen one or two of its creations floating about online. Want to find out more about Dall-E 2 and how it works? Keep reading.
What is Dall-E 2?
Dall-E 2 was created by artificial intelligence company Open AI. They describe it as an AI system that can create realistic images and art from text descriptions. Open AI originally introduced Dall-E in 2021 with Dall-E 2 arriving a year later, able to generate images with up to four times better resolution for a more accurate and realistic impression.
The name Dall-E 2 is a portmanteau of two figures: the iconic Spanish artist Salvador Dalí and Pixar’s own adorable robot Wall-E, giving credit to both the artistic and robotic inspirations of the software.
It works by users entering sentences that describe what they want to see. Dall-E 2 will create the image, and it carries the ability to combine different styles, concepts and attributes to forge the end result.
For example, the image below was created with the prompt, “An astronaut riding a horse in a photorealistic style”. The image below that used the prompt, “A bowl of soup that is a portal to another dimension as digital art”.
As you can see from these images, Dall-E 2 is capable of creating a wide range of photos, and it can even expand images beyond what was on the original canvas, shown here by the expansion of a classic painting, Girl with a Pearl Earring.
How does Dall-E 2 work?
Dall-E 2 is capable of understanding the relationship between the images and the text written to describe them through a process called Diffusion. This starts with a pattern of random dots and then it gradually changes that pattern into the specified image when it starts to recognise it.
To simplify the process, the AI is able to take the sentence prompt it is given and create a ‘mental’ image of what it should look like. Once this image has been created, it is able to take the caption and combine them into a real image with more detail. And since it can differentiate between art styles, such as modern or abstract, it can start pulling from different sources to create new images that have distinctive styles and looks.
Can I use Dall-E 2?
Originally, access to Dall-E 2 was limited and the waiting list was very long. Since September 2022 the queue to use it has gone, meaning anyone can use the software.
To sign up you can click on this link or visit the Open AI website. Once you have created an account, you can start creating images right away. For this image, we asked Dall-E 2 to create a “Mushroom with eyes and a face, digital art”.