← Back to Discover

DALL-E 3

OpenAI's text-to-image model, available in ChatGPT and through the API, built to follow the wording of a prompt closely with ChatGPT helping to write and refine prompts.

Image generationModel APIs
Visit website ↗

What it does

DALL·E 3 is an image generation model from OpenAI. The page says it is available to all ChatGPT users and to developers through the API. It is presented as an improvement over DALL·E 2 on the same prompts, with the aim of following your words closely so you do not have to learn special prompting tricks. The model is built natively on ChatGPT. You describe an idea in anything from a sentence to a paragraph, and ChatGPT writes a detailed prompt for the model. If an image is close but not right, you ask ChatGPT for a tweak in a few words. The page states that you may reprint, sell or merchandise the images you create without asking OpenAI for permission. The model is designed to decline requests that name a public figure or ask for the style of a living artist, and creators can opt their images out of training for future image models. OpenAI also says it is experimenting with a provenance classifier that can tell whether an image came from DALL·E 3.

Capabilities

  • Text-to-image generation that sticks closely to the prompt
  • ChatGPT writes and refines detailed prompts from a short idea
  • Ask ChatGPT for small tweaks to an image you almost like
  • Available to ChatGPT users and to developers through the API
  • Declines requests for public figures by name and for living artists' styles
  • Creators can opt their images out of future training

When to use it

  • You want images made from a written description inside ChatGPT
  • You are a developer adding image generation through OpenAI's API
  • You want to use generated images commercially, which the page says needs no permission from OpenAI