one year on
OpenAI unveils DALL-E 3 with native ChatGPT integration and artist opt-out for training data
The upgraded text-to-image model uses ChatGPT to generate and refine prompts, promising better text rendering and prompt adherence, while adding a mechanism for artists to block their work from future training.
OpenAI today announced DALL-E 3, the latest version of its text-to-image generator, which uses ChatGPT to help fill in prompts. Subscribers to ChatGPT Plus and ChatGPT Enterprise, as well as users of Bing Chat and Bing Chat Enterprise, can describe an image in natural language and let ChatGPT expand the prompt into a detailed description for DALL-E 3. The results appear inside the chat interface.
The company claims DALL-E 3 produces higher-quality images that better follow longer prompts and handle tricky elements like text and human hands. It also rejects requests to mimic living artists or depict public figures. A notable addition: artists can opt out of having their work used to train future versions of the model, a move that comes amid ongoing lawsuits from creators alleging unauthorized use of copyrighted material.
DALL-E 3 will roll out to premium ChatGPT users in October, with API access to follow. OpenAI has not announced a free web tool. The launch comes as the generative AI race heats up, particularly in image-synthesizing tools, with competitors like Midjourney and Stability AI continuing to refine their models.
One year later — open only if you can handle spoilers
DALL-E 3 effectively set a new standard for integration between language and image models. Competitors like Midjourney later added their own conversational interfaces, but OpenAI's head start with ChatGPT cemented its lead in the consumer multimodal space.