OpenAI Unveils DALL·E 3 Text-to-Image Model for ChatGPT

OpenAI has just announced its new DALL·E 3 advanced text-to-image model, which is set to launch in early October for ChatGPT Plus and Enterprise users.

DALL·E 3 represents a remarkable leap in the ability to generate images that precisely match the provided text. Even compared to its predecessor, DALL·E 2, DALL·E 3 showcases significant improvements.

One of the key highlights is that DALL·E 3 is seamlessly integrated into ChatGPT, allowing users to engage ChatGPT as a brainstorming partner and prompt refiner.

By simply sharing your idea with ChatGPT, it will craft tailored, detailed prompts for DALL·E 3, breathing life into your concept.

Furthermore, if an image isn’t quite perfect, a few words to ChatGPT can prompt tweaks to refine it.

Crucially, the images generated using DALL·E 3 remain the property of the users, with no requirement for OpenAI’s permission to reuse, sell, or merchandise them.

OpenAI, however, maintains its commitment to safety by implementing restrictions on DALL·E 3 to prevent the generation of violent, adult, or hateful content. Additionally, the model has been enhanced to decline requests that mention public figures by name, reducing potential misuse.

Collaborating with domain experts known as “red teamers,” OpenAI has improved the model’s safety performance, particularly in areas like generating public figures and mitigating biases associated with over/under-representation in visuals.

DALL·E 3 also possesses the ability to decline requests for images in the style of living artists, respecting the art community’s concerns. Furthermore, creators have the option to exclude their images from future image generation model training.