OpenAI updated the renderer according to the DALL-E text description

Nonprofit OpenAI has introduced a new version of its text annotation image generator that creates realistic images with higher resolution and lower latency than the original.

In DALL-E 2, users can select and edit specific areas of existing images, add or remove elements with their shadows, create collages and variations of finished drawings.

The neural network produces images with a resolution of 1024 pixels, which is four times higher than the original model.

The DALL-E service (a combination of the names of artist Salvador Dali and the cartoon character WALL-E) is based on the OpenAI CLIP computer vision model announced in 2021.

“He took the GPT-3 approach from the original model language and applied it to create an image: we put the images into a series of words and then learned to predict what would happen next,” said OpenAI researcher Prafulla Dhariwal.

Today, the DALL-E 2 is available to testers who have signed up for the waitlist. Users are not allowed to create obscene and extremist images, as well as materials related to “current geopolitical events”.

Recall that in January a machine learning engineer created a Pokemon generator based on the DALL-E model.

In August 2021, an enthusiast developed a scaled-down version of OpenAI’s Text Annotation Drawing Generator.

Follow ForkLog on TikTok!