The image generator within the AI-powered Bing Chat gets a big update today: Microsoft announced that OpenAI’s latest DALL-E 3 model is Now available to all Bing Chat and Bing Image Creator users. It launched in the last week or so, first for Bing Enterprise users and then for Bing Image Creator, but is now open to everyone.
Bing will get DALL-E 3 access even before OpenAI’s ChatGPT: this is expected this month, but only for paying users. Microsoft will probably be the most popular image generation tool for a while.
DALL-E 3 is, of course, the third version of OpenAI’s image generation model. The company claims that it understands the instructions much better than before and can create more creative and more photorealistic images. It’s also designed to be much easier to use; DALL-E 3 is integrated into Bing Chat and ChatGPT rather than powering a standalone product, so you can create and refine your image by conversing with a chatbot instead of endlessly trying to refine your initial suggestion.
OpenAI has also integrated new security tools into DALL-E 3: it’s designed not to recreate images of public figures, for example, and it shouldn’t create hateful or NSFW images either. Within Bing Image Creator, Microsoft is also embedding watermarks into each image to identify them as AI-generated and has created its own content moderation system. As always, though, the proof will be in the pictures.
Microsoft plans to use DALL-E technology not only in Bing. It’s working on an AI image creation tool in the Paint app called Paint Cocreator, for example, which will bring the DALL-E model directly into Windows.
Anyone can theoretically use DALL-E 3 now via Bing, although every time I’ve tried so far, it’s been “unable to process new requests” and refused to create my image. Apparently the servers have been overloaded for a while: Microsoft’s Published Mikhail Parakhin“We expected strong interest, but we didn’t expect this much.”
When you create an image, it will not only create the one you requested, but will also offer suggestions on where to go next: “Can you add a rainbow in the background?” “Make him a cat instead of a dog.” “Add some birds around the waterfall.” (Bing Chat wouldn’t create those images for me either, because the system was overloaded, but it seems like a fun way to create alongside the tool.)