Have you ever wanted to describe an image hand ave it created for you? With Open AI’s new artificial intelligence system DALL·E 2 you can do just that, as it creates realistic images and art from a worded description.
Not only can DALL·E 2 create images from the text, but it can also combine concepts, attributes and artistic styles. While some of the best photo editing software can be used to achieve the same results, DALL·E 2 requires a lot less effort from the creator. It can add and remove elements into a photo while taking into account natural elements such as shadows, reflections and texture. Are you wondering, like us, how AI is changing photography and photo editing?
The company Open AI first introduced DALL·E in January 2021, but the new and improved version is able to create higher resolution images, better comprehension and new capabilities. DALL·E was created using a neural network that trained it to recognize pictures from their text descriptions. A neural network is one method used in AI to teach computers to process data in a way that is inspired by the human brain. Using deep learning, DALL·E 2 is able to recognize individual objects and combine them into a picture in a way that looks realistic. I recently tried it out...
For example, if you asked DALL·E 2 to create an image of a koala riding a motorbike, it would source an image of the individual objects but combine them to give a photo-realistic end result.
AI technology is constantly evolving and DALL·E 2 helps us to understand how AI systems see our world. It does however have its limitations, and human error can cause the system to not work as it should.
If pictures are labeled with the wrong text, it would get confused and generate the wrong images. There are also gaps in the DALL·E 2’s training which means it might generate what it thinks you mean – rather than what you actually mean.
Generally, DALL·E 2 will work best for things that are very specific. So long as it has learned what an animal or object is, it will be able to generate images for it. Using the example of a baboon (because why not) DALL·E 2 could generate hundreds of images, but if you typed howler monkey and it doesn’t know what it is, it will generate what it thinks is correct – like a howling monkey.
You can join the waitlist to access DALL·E 2 and over time invites will be sent out. Having just spent a bit of time playing with what is available on the website, I am already excited to see what exactly DALL·E 2 is capable of. Will this be the death of composite photography or will it be a way to quickly imagine a scene before creating it “properly” in photo editing software? Either way, it'll be fun to see how creative and whacky you can get with it.