Dall-E 3

Dall-E 3 is a text-to-image artificial intelligence (AI) model developed by OpenAI. The model takes natural language prompts and generates text outputs. Announced on September 20, 2023, Dall-E 3 is the successor to OpenAI's previous text-to-image model, Dall-E 2. Dall-E 3 is integrated with ChatGPT, and users can use the chatbot to generate detailed prompts to guide the model. OpenAI describes ChatGPT as a "brainstorming partner," refining prompts and converting simple user-provided sentences into detailed paragraphs to improve the accuracy of Dall-E 3.

OpenAI's Dall-E 3 introduction video showcasing some of it's new capabilities.

Initially in research preview following the announcement on September 20, Dall-E 3 became available in ChatGPT Plus and Enterprise on October 19, 2023, with plans to make the model available via the API and in Labs later in 2023.

Dall-E 3 is the culmination of numerous research advancements from within OpenAI and outside of the company. OpenAI states that Dall-E 3 understands significantly more nuance and detail than their previous systems, allowing users to translate ideas into more accurate images. OpenAI goes on to state that Dall-E 3 creates images that are more visually striking and crisper in detail. In particular, Dall-E 3 can reliably render intricate details such as text, hands, and faces, as well as support both landscape and portrait aspect ratios.

Comparison of Dall-E 2 and Dall-E 3 outputs given the same prompt.

Dall-E 3's improved capabilities were achieved by training an image captioner to generate better textual descriptions for images. These images with improved captions were then used to train Dall-E 3, producing a model more responsive to user-supplied captions. Dall-E 3 is described in greater detail in the system card, released by OpenAI on October 3, 2023.

Users can freely use images generated using Dall-E 3 without permission for reprinting, selling, and merchandising. OpenAI has introduced additional safety options for Dall-E 3. The model will decline requests that ask for public figures by name or ask for an image in the style of a living artist. The company is also experimenting with a provenance classifier, an internal tool that helps identify whether an image was generated using Dall-E 3.