Check out that gorgeously detailed image above. It was created using DALL-E 3 through the Bing Image Creator with the prompt “a Komodo dragon laying on a plush deep green pillow on the floor of a mansion.” Microsoft just announced that DALL-E 3 is generally available for free in Bing Chat or with the Bing Image Generator.
Created by OpenAI, DALL-E 3 is the company’s latest version of a text-to-image model and promises to deliver improved detail in images and greater accuracy in faces, text, and even human hands.
Also: Google Assistant is finally getting the AI upgrades it deserves. Here’s what’s new
Microsoft has been using what it referred to as a more advanced version of DALL-E 2 in Bing Chat and the Bing Image Creator, but today’s announcement appears to confirm that it was not using DALL-E 3 until now. Even though images created with Bing were superior to those created directly with DALL-E 2, today’s upgrade is evident in the image output’s improved quality, as shown below.
The images on the left were created with the Bing Image Creator before integrating DALL-E 3, while the ones on the right were created with the same prompt after Microsoft made its announcement. Before the update, the creature looked like a stuffed animal on a scooter. With DALL-E 3, the images on the right show greater detail and are more photorealistic.
Also: Six skills you need to become an AI prompt engineer
OpenAI, also the creator of ChatGPT, improved the aesthetics of the images the model puts out, as well as the coherence and prompt following. These changes make the DALL-E 3-generated images more logically consistent with the prompts given, so users can provide more detail than before in their prompts without fear of confusing the model.
Aside from producing images that are more accurate to the prompts, Microsoft recently announced that images created with Bing Image Creator and Bing Chat will include an invisible digital watermark with the date and time it was created to confirm that the image was AI-generated.
More on AI tools