OpenAI has introduced a new version of its image editor called DALL-E 3

OpenAI has unveiled a new version of its DALL-E image generator and announced plans to integrate it into ChatGPT. The developers promise that DALL-E 3 will better understand prompts and analyze complex instructions without mixing different query elements, and will generate "extremely detailed and accurate images." Users have also been given the ability to compose prompts with ChatGPT.

In the description of the neural network, OpenAI assures that DALL-E 3 significantly outperforms its predecessor in generating images from complex prompts. The model is able to accurately represent scenes with specific objects and take into account their interrelationships. DALL-E 3 is better able to handle subtleties such as human hands, which is a challenging task for artificial intelligence. Even with the same queries, the results of DALL-E 3 are significantly superior to those of DALL-E 2, the developers note.

DALL-E 3 provides the ability to use ChatGPT as a partner for brainstorming and refining promts. Users can request ChatGPT to compile a detailed promt and make changes to the resulting image if it does not fully comply with the request.

DALL-E 3 is currently in an exploratory pre-release stage and will be available to ChatGPT Plus and Enterprise subscribers in October via API.

OpenAI specifies that, like DALL-E 2, the images generated with DALL-E 3 are owned by the user and they do not need permission to reprint, sell, or distribute them.

It is important to note that DALL-E 3 includes features that limit the generation of certain images. For example, the system will reject requests containing the names of public figures or "harmful bias" to reduce the risk of spreading propaganda and misinformation. The developers are also testing an origin classifier that will help determine whether an image was created by a neural network. In addition, DALL-E will not generate images in the style of contemporary living and creating artists.