OpenAI officially announced DALL·E 3, a modern text-to-image system built natively on ChatGPT to generate artwork by simply talking to the chatbot.
OpenAI officially announced DALL·E 3, a modern text-to-image system built natively on ChatGPT to generate artwork by simply talking to the chatbot.
Generated with DALL-E 3: “A vibrant yellow banana-shaped couch sits in a cozy living room, its curve cradling a pile of colorful cushions. on the wooden floor, a patterned rug adds a touch of eclectic charm, and a potted plant sits in the corner, reaching towards the sunlight filtering through the window.”, Image by OpenAI
What innovation does DALL-E 3 bring?
We gathered all the information shared in OpenAI’s announcement about the latest version of DALL-E, and boy, it sounds promising!
No need to be a prompt engineering grand master – DALL-E 3 enables you to use the ChatGPT conversational interface to improve the images you generate. This means that if you didn’t like what it produced, you can simply talk with ChatGPT and ask for the changes you’d like to make. This removes the complexity associated with prompt engineering, which requires you to iterate over the prompt.From October, DALL-E 3 will be available through ChatGPT and API for those with the Plus or Enterprise version.Major improvement in the quality of products compared to DALL-E 2. This is a very vague statement provided by OpenAI, which is also hard to measure, but personally, they haven’t failed me so far, so I’m really excited to see the results.DALL-E 2 Vs. DALL-E 3, image by OpenAIAs with DALL-E 2, the images created with DALL-E 3 don’t have copyrights and you don’t need OpenAI’s permission to reprint, sell or merchandise them.Requesting images in the style of a living artist will not be available, and they can even ask not to be included in the future training process.Using names of familiar characters is no longer allowed and they improved safety performances in the from of harmful biases related to visual over/under-representation.OpenAI are working on a new tool to detect when an image was created with AI. They state that the main goal is to detect what AI-generated images are used for. This sounds very ambitious and might work in some cases, but I doubt if it will overcome SOTA deep fake models.
Let us know what you think about the release notes and if they’ve lived up for your expectation 👉 Discord
Generated with DALL-E 3: “Photo of a lychee-inspired spherical chair, with a bumpy white exterior and plush interior, set against a tropical wallpaper.”, image by OpenAI
Resource: OpenAI announcement about DALL-E 3
