(ORDO NEWS) — In January 2021, the OpenAI Artificial Intelligence Lab created DALL-E, a neural network that generates cartoony images from text captions.
Now the company has improved the program – it has learned to make better and more detailed images.
Some of the images are truly amazing. DALL-E 2 is an updated neural network from OpenAI that can generate more detailed pictures from a text description.
For example, the description “astronaut” , “riding a horse” and “in a photorealistic style” resulted in the generation of this image:
The most striking thing is that the updated neural network independently understands how and in what position to put an astronaut on a horse and knows what photorealism is.
According to the OpenAI website , the neural network can not only generate images, but also edit them. Upon request, she can change some of the details of the photo by adding or removing specific elements.
And this is the monster yarn bowl also created by DALL E 2:
How does DALL-E work?
According to the description, “DALL·E 2 explored the relationship between images and the text used to describe them.
It uses a process called “scattering” that starts with a pattern of random dots and gradually changes this pattern towards the image when the neural network recognizes certain aspects of this image .
So far, only selected users can test the neural network. However, in the future, OpenAI plans to add a neural network to the API so that other developers can create applications using it.
Contact us: [email protected]ws.com